Add files using upload-large-folder tool

.gitattributes

@@ -4939,3 +4939,201 @@ parse/train/rklMnyBtPB/rklMnyBtPB_origin.pdf filter=lfs diff=lfs merge=lfs -text
 parse/train/S1x4ghC9tQ/S1x4ghC9tQ_origin.pdf filter=lfs diff=lfs merge=lfs -text
 parse/train/S1x4ghC9tQ/S1x4ghC9tQ_span.pdf filter=lfs diff=lfs merge=lfs -text
 parse/train/S1x4ghC9tQ/S1x4ghC9tQ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/zQTezqCCtNx/zQTezqCCtNx_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/zQTezqCCtNx/zQTezqCCtNx_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByexElSYDr/ByexElSYDr_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByexElSYDr/ByexElSYDr_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByexElSYDr/ByexElSYDr_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJWGdbbCW/HJWGdbbCW_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJWGdbbCW/HJWGdbbCW_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJWGdbbCW/HJWGdbbCW_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkGabzZgl/rkGabzZgl_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkGabzZgl/rkGabzZgl_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkGabzZgl/rkGabzZgl_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJg3zxBYwH/rJg3zxBYwH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJg3zxBYwH/rJg3zxBYwH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJg3zxBYwH/rJg3zxBYwH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xSperKvH/B1xSperKvH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xSperKvH/B1xSperKvH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xSperKvH/B1xSperKvH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/oyZxhRI2RiE/oyZxhRI2RiE_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/oyZxhRI2RiE/oyZxhRI2RiE_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/oyZxhRI2RiE/oyZxhRI2RiE_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/S1zz2i0cY7/S1zz2i0cY7_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/S1zz2i0cY7/S1zz2i0cY7_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/S1zz2i0cY7/S1zz2i0cY7_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B13njo1R-/B13njo1R-_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B13njo1R-/B13njo1R-_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B13njo1R-/B13njo1R-_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ry018WZAZ/ry018WZAZ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ry018WZAZ/ry018WZAZ_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ry018WZAZ/ry018WZAZ_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SyK00v5xx/SyK00v5xx_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SyK00v5xx/SyK00v5xx_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SyK00v5xx/SyK00v5xx_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByJHuTgA-/ByJHuTgA-_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByJHuTgA-/ByJHuTgA-_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByJHuTgA-/ByJHuTgA-_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJx-3grYDB/HJx-3grYDB_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJx-3grYDB/HJx-3grYDB_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJx-3grYDB/HJx-3grYDB_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkSDMA36Z/BkSDMA36Z_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkSDMA36Z/BkSDMA36Z_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkSDMA36Z/BkSDMA36Z_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJg_ECEKDr/HJg_ECEKDr_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJg_ECEKDr/HJg_ECEKDr_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJg_ECEKDr/HJg_ECEKDr_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/pULTvw9X313/pULTvw9X313_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/pULTvw9X313/pULTvw9X313_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/pULTvw9X313/pULTvw9X313_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByeSdsC9Km/ByeSdsC9Km_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByeSdsC9Km/ByeSdsC9Km_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByeSdsC9Km/ByeSdsC9Km_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r154_g-Rb/r154_g-Rb_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r154_g-Rb/r154_g-Rb_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r154_g-Rb/r154_g-Rb_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJeD3CEFPH/SJeD3CEFPH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJeD3CEFPH/SJeD3CEFPH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJeD3CEFPH/SJeD3CEFPH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1gL-2A9Ym/H1gL-2A9Ym_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1gL-2A9Ym/H1gL-2A9Ym_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1gL-2A9Ym/H1gL-2A9Ym_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HygjqjR9Km/HygjqjR9Km_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HygjqjR9Km/HygjqjR9Km_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HygjqjR9Km/HygjqjR9Km_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1vEXaxA-/H1vEXaxA-_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1vEXaxA-/H1vEXaxA-_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1vEXaxA-/H1vEXaxA-_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/MDsQkFP1Aw/MDsQkFP1Aw_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/MDsQkFP1Aw/MDsQkFP1Aw_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/MDsQkFP1Aw/MDsQkFP1Aw_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/CGFN_nV1ql/CGFN_nV1ql_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/CGFN_nV1ql/CGFN_nV1ql_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/CGFN_nV1ql/CGFN_nV1ql_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByC7ww9le/ByC7ww9le_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByC7ww9le/ByC7ww9le_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByC7ww9le/ByC7ww9le_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/3hGNqpI4WS/3hGNqpI4WS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/3hGNqpI4WS/3hGNqpI4WS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/3hGNqpI4WS/3hGNqpI4WS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJeU_1SFvr/rJeU_1SFvr_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJeU_1SFvr/rJeU_1SFvr_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJeU_1SFvr/rJeU_1SFvr_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/JG-SlCAx5_K/JG-SlCAx5_K_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/JG-SlCAx5_K/JG-SlCAx5_K_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/JG-SlCAx5_K/JG-SlCAx5_K_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/te7PVH1sPxJ/te7PVH1sPxJ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/te7PVH1sPxJ/te7PVH1sPxJ_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/te7PVH1sPxJ/te7PVH1sPxJ_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkeHuCVFDr/SkeHuCVFDr_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkeHuCVFDr/SkeHuCVFDr_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkeHuCVFDr/SkeHuCVFDr_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/PrzjugOsDeE/PrzjugOsDeE_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/PrzjugOsDeE/PrzjugOsDeE_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/PrzjugOsDeE/PrzjugOsDeE_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/bw5Arp3O3eY/bw5Arp3O3eY_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/bw5Arp3O3eY/bw5Arp3O3eY_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/bw5Arp3O3eY/bw5Arp3O3eY_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rywDjg-RW/rywDjg-RW_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rywDjg-RW/rywDjg-RW_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rywDjg-RW/rywDjg-RW_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJz6tiCqYm/HJz6tiCqYm_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJz6tiCqYm/HJz6tiCqYm_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJz6tiCqYm/HJz6tiCqYm_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/eQe8DEWNN2W/eQe8DEWNN2W_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/eQe8DEWNN2W/eQe8DEWNN2W_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/eQe8DEWNN2W/eQe8DEWNN2W_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJtfOEn6-/SJtfOEn6-_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJtfOEn6-/SJtfOEn6-_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJtfOEn6-/SJtfOEn6-_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJXTf9Bxg/rJXTf9Bxg_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJXTf9Bxg/rJXTf9Bxg_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rJXTf9Bxg/rJXTf9Bxg_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/OP4ePyQXNu-da/OP4ePyQXNu-da_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/OP4ePyQXNu-da/OP4ePyQXNu-da_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/OP4ePyQXNu-da/OP4ePyQXNu-da_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylR6n4tPS/SylR6n4tPS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylR6n4tPS/SylR6n4tPS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylR6n4tPS/SylR6n4tPS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Byl8hhNYPS/Byl8hhNYPS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Byl8hhNYPS/Byl8hhNYPS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Byl8hhNYPS/Byl8hhNYPS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/x5hh6N9bUUb/x5hh6N9bUUb_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/x5hh6N9bUUb/x5hh6N9bUUb_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/x5hh6N9bUUb/x5hh6N9bUUb_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJgEl3A5tm/SJgEl3A5tm_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJgEl3A5tm/SJgEl3A5tm_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJgEl3A5tm/SJgEl3A5tm_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkJsCIcgl/BkJsCIcgl_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkJsCIcgl/BkJsCIcgl_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/BkJsCIcgl/BkJsCIcgl_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByMVTsR5KQ/ByMVTsR5KQ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByMVTsR5KQ/ByMVTsR5KQ_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/ByMVTsR5KQ/ByMVTsR5KQ_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/xTJEN-ggl1b/xTJEN-ggl1b_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/xTJEN-ggl1b/xTJEN-ggl1b_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/xTJEN-ggl1b/xTJEN-ggl1b_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/blRJEZfyem/blRJEZfyem_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/blRJEZfyem/blRJEZfyem_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/blRJEZfyem/blRJEZfyem_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Ua6zuk0WRH/Ua6zuk0WRH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Ua6zuk0WRH/Ua6zuk0WRH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/Ua6zuk0WRH/Ua6zuk0WRH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/A7pvvrlv68/A7pvvrlv68_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/A7pvvrlv68/A7pvvrlv68_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/A7pvvrlv68/A7pvvrlv68_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xGGTEtDH/B1xGGTEtDH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xGGTEtDH/B1xGGTEtDH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/B1xGGTEtDH/B1xGGTEtDH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJUOHGWRb/HJUOHGWRb_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJUOHGWRb/HJUOHGWRb_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJUOHGWRb/HJUOHGWRb_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJlpM3RqKQ/SJlpM3RqKQ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJlpM3RqKQ/SJlpM3RqKQ_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJlpM3RqKQ/SJlpM3RqKQ_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/yehlf2AvSD_/yehlf2AvSD__span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/yehlf2AvSD_/yehlf2AvSD__origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/yehlf2AvSD_/yehlf2AvSD__layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HyQJ-mclg/HyQJ-mclg_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HyQJ-mclg/HyQJ-mclg_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HyQJ-mclg/HyQJ-mclg_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gixp4FPH/r1gixp4FPH_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gixp4FPH/r1gixp4FPH_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gixp4FPH/r1gixp4FPH_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/X4y_10OX-hX/X4y_10OX-hX_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/X4y_10OX-hX/X4y_10OX-hX_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/X4y_10OX-hX/X4y_10OX-hX_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/p99rWde9fVJ/p99rWde9fVJ_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/p99rWde9fVJ/p99rWde9fVJ_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/p99rWde9fVJ/p99rWde9fVJ_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rygMWT4twS/rygMWT4twS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rygMWT4twS/rygMWT4twS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rygMWT4twS/rygMWT4twS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/spjlJ4jeM_/spjlJ4jeM__layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/spjlJ4jeM_/spjlJ4jeM__origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/spjlJ4jeM_/spjlJ4jeM__span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1lK_lBtvS/H1lK_lBtvS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1lK_lBtvS/H1lK_lBtvS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/H1lK_lBtvS/H1lK_lBtvS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/T3kmOP_cMFB/T3kmOP_cMFB_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/T3kmOP_cMFB/T3kmOP_cMFB_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/T3kmOP_cMFB/T3kmOP_cMFB_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkgbmyHFDS/SkgbmyHFDS_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkgbmyHFDS/SkgbmyHFDS_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SkgbmyHFDS/SkgbmyHFDS_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkl3m1BFDB/rkl3m1BFDB_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkl3m1BFDB/rkl3m1BFDB_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/rkl3m1BFDB/rkl3m1BFDB_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/6Tm1mposlrM/6Tm1mposlrM_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/6Tm1mposlrM/6Tm1mposlrM_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/6Tm1mposlrM/6Tm1mposlrM_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylzhkBtDB/SylzhkBtDB_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylzhkBtDB/SylzhkBtDB_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SylzhkBtDB/SylzhkBtDB_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJNRHiAcYX/SJNRHiAcYX_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJNRHiAcYX/SJNRHiAcYX_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/SJNRHiAcYX/SJNRHiAcYX_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gBOxSFwr/r1gBOxSFwr_layout.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gBOxSFwr/r1gBOxSFwr_origin.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/r1gBOxSFwr/r1gBOxSFwr_span.pdf filter=lfs diff=lfs merge=lfs -text
+parse/train/HJgXCV9xx/HJgXCV9xx_origin.pdf filter=lfs diff=lfs merge=lfs -text

parse/dev/RzXb6a3H3rs/RzXb6a3H3rs.md

@@ -0,0 +1,290 @@
+# LEARNING TO PROMPT FOR CONTINUAL LEARNING
+
+Anonymous authors Paper under double-blind review
+
+# ABSTRACT
+
+The mainstream learning paradigm behind continual learning has been to adapt the model parameters to non-stationary data distributions, where catastrophic forgetting is the central challenge. This work explores a new paradigm for continual learning – learning to dynamically prompt the model to learn tasks sequentially under different task transitions. Specifically, our method, Learning to Prompt for Continual Learning (L2P), prepends a subset of learnable parameters (called Prompts) from a larger set (called Prompt Pool) to the input embeddings. The training objective is designed to dynamically select and update prompts from the prompt pool to learn tasks sequentially given a pretrained backbone model. Under our new framework, instead of mitigating catastrophic forgetting via adapting large model parameters as in the previous continual learning paradigm, we tackle the problem of learning better small prompt parameters. In this framework, the prompt pool explicitly manages task-invariant and task-specific knowledge while maintaining model plasticity. The proposed L2P outperforms previous work in terms of forgetting on all datasets, including rehearsal-based methods on certain benchmarks, with privacy benefits from not requiring access to the data of previous tasks. Moreover, when L2P is additionally equipped with a rehearsal buffer, it matches the performance of training all tasks together, which is often regarded as an upper bound in continual learning. Source code will be released.
+
+# 1 INTRODUCTION
+
+Contrary to ordinary supervised learning that trains on independent and identically distributed (i.i.d.) data, continual learning tackles the problem of training a single model on non-stationary data distributions where different classification tasks are presented sequentially. Mainstream continual learning methods (Parisi et al., 2019; Mai et al., 2021) follow a natural learning paradigm: adapting the entire model continually as the data distribution shifts. However, since the model only has access to the data in an individual phase of the learning cycle, it is prone to overfit on the currently available data and suffers from performance deterioration on the previously trained data. This is commonly known as catastrophic forgetting (McCloskey & Cohen, 1989).
+
+In addition to the catastrophic forgetting problem, other challenges in continual learning have recently been receiving increasing attention (Hadsell et al., 2020): (1) knowledge transfer: the model should be able to transfer knowledge between tasks by identifying shared knowledge among tasks; (2) model plasticity: the model should be able to keep learning new tasks effectively by capturing task-specific knowledge; and (3) task-agnosticity: it is desirable that a continual learning algorithm can handle the case where distribution shifts gradually without clear task boundaries.
+
+On the other hand, prompt-based learning, or prompting, has recently achieved great success in the field of natural language processing (NLP) as a new transfer learning technique (Liu et al., 2021). Prompting techniques design model inputs with textual prompt tokens containing additional taskspecific information, such that the pretrained language model can process parameterized inputs in order to perform prompt-specific prediction. Several methods (Lester et al., 2021; Shin et al., 2020; Li & Liang, 2021) further make prompts learnable to allow the overall backbone model to extract task-specific information automatically. Intuitively, prompt-based learning reformulates learning downstream tasks from directly adapting model weights to designing prompts that enable the model perform tasks conditionally. A prompt encodes task-specific knowledge and has the ability to utilize pre-trained frozen models more effectively than ordinary fine-tuning (Lester et al., 2021; Raffel et al., 2020). Inspired by these recent advances in prompt learning, we revisit continual learning from a different perspective: Can we encode task-specific information of continual tasks into a shared parameterized prompt space in order to allow a pre-trained model to perform conditional prediction during the continual learning process?
+
+![](images/b2b885e90ab9fc48f9e4bc5fae2bc06fc57b7771b4a1d50e0dbb70dda6b510b5.jpg)  
+Figure 1: Overview of the L2P framework. Compared with typical continual learning methods (left) that adapt model weights to tasks sequentially, L2P (right) uses a single backbone model and learns a prompt pool to adapt tasks.
+
+To this end, we propose a new continual learning method called Learning to Prompt for Continual Learning (L2P). Figure 1 gives an overview of our method and demonstrates how it differs from typical continual learning methods. L2P leverages the representative features from pretrained models; however, instead of tuning the parameters during the continual learning process, L2P keeps the pretrained model untouched, and instead learns a set of prompts that dynamically help models solve corresponding tasks, thus mitigating catastrophic forgetting. The prompts are structured in a key-value shared memory space called the prompt pool, and we design a query mechanism to dynamically lookup a subset of task-relevant prompts based on the instance-wise input features. The prompt pool, which is optimized jointly with the supervised loss, ensures that shared prompts encode shared knowledge for knowledge transfer, and unshared prompts encode task-specific knowledge that help maintain model plasticity. The instance-wise query mechanism removes the necessity of knowing the task identity or boundaries, enabling task-agnostic continual learning. The selected prompts are then prepended to the input embeddings (Figure 2), which implicitly add task-relevant guidance to pretrained models, so that the model can use the most useful pretrained features to conduct corresponding tasks. In summary, this work makes the following contributions:
+
+1. We propose a novel method, called L2P, that addresses multiple challenges in continual learning: (1) we leverage pretrained models and prompting techniques to mitigate catastrophic forgetting; (2) we design a novel key-value paired prompt pool to achieve knowledge sharing and maintain model plasticity; and (3) we devise an instance-wise query mechanism to enable task-agnostic learning.   
+2. We conduct comprehensive experiments to demonstrate the effectiveness of L2P on multiple continual learning benchmarks, including class-incremental, task-agnostic, and domainincremental settings. The proposed L2P outperforms previous works in terms of forgetting on all datasets, beating rehearsal based methods on certain benchmarks and providing practical advantages over them by avoiding privacy issues of task data sharing present in some applications (Delange et al., 2021). Moreover, when equipped with a rehearsal buffer in applications with less strict privacy constraints, L2P matches the performance of training all tasks together, which is often regarded as an upper bound in continual learning.   
+3. To the best of our knowledge, we are the first to introduce the idea of prompting in the field of continual learning to address some of the key challenges in continual learning.
+
+# 2 RELATED WORK
+
+Continual learning. There are three main categories of recent continual learning algorithms: Regularization-based methods (Kirkpatrick et al., 2017; Zenke et al., 2017; Li & Hoiem, 2017; Aljundi et al., 2018) limit the plasticity of the model by limiting the learning rate on important parameters for previous tasks. Although these methods address catastrophic forgetting to some extent, they cannot get satisfactory performance under more challenging settings, e.g., class-incremental setting (Mai et al., 2021). Rehearsal-based methods (Chaudhry et al., 2018; 2019; Hayes et al., 2019) construct a buffer to save samples from older tasks to train with data from the current task. These methods are state-of-the-art on various benchmarks (Parisi et al., 2019; Mai et al., 2021). However, rehearsal-based methods are not applicable to scenarios where data privacy should be taken into account (Shokri & Shmatikov, 2015). Architecture-based methods either expand the network (Rusu et al., 2016; Yoon et al., 2017) or prune the network (Mallya & Lazebnik, 2018; Wang et al., 2020). The former suffers from scalability issue as parameters scale up linearly with the number of tasks, and the latter are sensitive to hyperparameters.
+
+Prompting. Prompting, or prompt-based learning, has been widely explored in the field of natural language processing (Kumar et al., 2016; McCann et al., 2018; Radford et al., 2019; Schick & Schutze ¨ , 2020). The high-level idea of prompting is to apply a function to modify the input text, so that the language model gets additional information about the task. However, the design of a prompting function is challenging and requires heuristics. Recent work, including prompt tuning (Lester et al., 2021) and prefix tuning (Li & Liang, 2021), seek to address this problem by applying learnable prompts in a continuous space, achieving satisfactory performance on transfer learning for pretrained language models. Nevertheless, to the best of our knowledge, the idea of prompting has never been studied systematically in continual learning.
+
+# 3 PREREQUISITES
+
+# 3.1 CONTINUAL LEARNING PROTOCOLS
+
+Continual learning is usually defined as training machine learning models on non-stationary data from sequential tasks. We define a sequence of tasks $\mathcal { D } = \{ \mathcal { D } _ { 1 } , \cdot \cdot \cdot , \mathcal { D } _ { T } \}$ , where the $t$ -th task $\mathcal { D } _ { t } = \{ ( \dot { \mathbf { x } } _ { i } ^ { t } , y _ { i } ^ { t } ) \} _ { i = 1 } ^ { n _ { t } }$ contains tuples of the input sample $\boldsymbol { x } _ { i } ^ { t } \in \mathcal { X }$ and its corresponding label $y _ { i } ^ { t } \in \mathcal { V }$ . The goal is to train a single model $f _ { \theta } : \mathcal { X }  \mathcal { Y }$ parameterized by $\theta$ , such that it predicts the label $y = f _ { \boldsymbol { \theta } } ( \pmb { x } ) \in \mathcal { y }$ given an unseen test sample $_ { \textbf { \em x } }$ from arbitrary tasks. Data from the previous tasks may not be seen anymore when training future tasks.
+
+Depending on the task transition environment, continual learning can be categorized into multiple settings with slightly different challenges. The common task, class, and domain incremental setting assumes task data $\mathcal { D } _ { t }$ arrives in sequence $t = \{ 1 , . . . , T \}$ in a discrete manner. Task-incremental assumes task identity is known at test time while class-incremental does not. Different from the task and class incremental settings where each task has different classes, domain-incremental learning maintains the same set of classes for every task and only changes the distribution of $_ { \textbf { \em x } }$ by task. In the more challenging task-agnostic setting, task data in $\mathcal { D }$ changes smoothly, and the task identity $t$ is unknown. Our paper tackles the more challenging class-incremental, task-agnostic, and domainincremental settings.
+
+# 3.2 PROMPT-BASED LEARNING AND BASELINES
+
+Prompt-based learning is an emerging technique in NLP. In contrast to traditional supervised finetuning, this type of methods design task-specific prompt functions to enable pre-trained models perform corresponding tasks (Liu et al., 2021). One of recent techniques, Prompt Tuning (PT) (Lester et al., 2021), proposes to simply condition frozen T5-like language models (Raffel et al., 2020) to perform down-streaming NLP tasks by learning prompt parameters that are prepended to the input tokens. While prompt-based learning has demonstrated success in NLP, to the best of our knowledge, the related research in computer vision and its application to continual learning remains under-investigated. Without loss of generality, here we introduce the definition of PT using the image modality given vision transformer-based models (Dosovitskiy et al., 2021; Vaswani et al., 2017). The definition is easy to generalize to other modalities and sequence-based models.
+
+Given an input of 2D image $\pmb { x } \in \mathbb { R } ^ { H \times W \times C }$ and a pretrained ViT (excluding the classification head) $f ~ = ~ f _ { r } \circ f _ { e }$ , where $f _ { e }$ is the input embedding layer, and $f _ { r }$ represents a stack of selfattention layers (Dosovitskiy et al., 2021). Images are reshaped to a sequence of flattened 2D patches $\pmb { x } _ { p } \in \mathbb { R } ^ { L \times ( S ^ { 2 } \cdot C ) }$ , where $L$ is the token length, i.e., the number of patches, $S$ is the patch size and $C$ is the original number of channels. To simplify notation, we assume the first token in $\scriptstyle { \pmb { x } } _ { p }$ is the [class] token as part of pre-trained model (Dosovitskiy et al., 2021). The pretrained embedding layer $f _ { e } : \mathbb { R } ^ { L \times ( S ^ { 2 } \cdot C ) }  \mathbb { R } ^ { L \times D }$ projects the patched image to the embedding feature $\pmb { x } _ { e } = f _ { e } ( x ) \in \mathbb { R } ^ { L \times D }$ , where $D$ is the embedding dimension. When solving multiple downstreaming tasks, we keep the large-scale pre-trained backbone frozen to maintain its generality following PT. The direct application of PT is to prepend learnable parameters $\boldsymbol { P _ { e } } \in \mathbb { R } ^ { L _ { p } \times D }$ , called a prompt, to the embedding feature $\pmb { x } _ { p } = [ P _ { e } ; \pmb { x } _ { e } ]$ , and feed the extended sequences to the model function $f _ { r } ( { \pmb x } _ { p } )$ for performing classification tasks. Different tasks have independent prompts and share one copy of the large model.
+
+![](images/4bada4c7fb75fc17f5d671b20d278b882ba3f46caedbd0ca7570db1b6c34be84.jpg)  
+Figure 2: The illustration of L2P at test time. During training time, we follow the same procedure and optimize the model as described in Section 4.3.
+
+Compared with ordinary fine-tuning classification heads with a fixed backbone, literature shows that prompt-based learning results in a sequence-based model with higher capacity to learn features (Liu et al., 2021; Lester et al., 2021). PT can be applied to task-incremental continual learning by learning independent prompts for each task. However, in more challenging settings when no task identity is available, choosing a prompt is more difficult.
+
+# 4 LEARNING TO PROMPT
+
+Our proposed method, Learning to Prompt for Continual Learning (L2P) is depicted in Figure 2. First, we select a subset of prompts from a key-value pair prompt pool based on our proposed instance-wise query mechanism. We then prepend the selected prompts to the input embedding. Finally, we feed the extended input embedding to the model, and optimize the classification loss and the prompt pool jointly. In the remainder of this section, we will introduce the critical designs of our method in detail, and discuss how L2P mitigates catastrophic forgetting and addresses some of the other challenges in continual learning (Hadsell et al., 2020), and describe the training procedure.
+
+# 4.1 FROM PROMPT TO PROMPT POOL
+
+The motivations of introducing prompt pool are threefold. First, the task index at test time is unknown so training task-independent prompts is not feasible. Second, even if the task-independent prompt can be known at test time, it prevents possible knowledge sharing between similar tasks (Hadsell et al., 2020). Third, while the simple way of learning a single shared prompt for all tasks enables knowledge sharing, it is challenging when tasks are diverse (see Section 5.3). Ideally one would learn a model that is able to share knowledge when tasks are similar, while maintaining knowledge independence otherwise. Thus, we propose using a prompt pool to store encoded knowledge, which can be flexibly grouped as an input to the model. The prompt pool is defined as
+
+$$
+\mathbf { P } = \{ P _ { 1 } , P _ { 2 } , \cdot \cdot \cdot , P _ { M } \} , \quad M = \mathrm { t o t a l n u m b e r o f p r o m p t s } ,
+$$
+
+where $P _ { j } \in \mathbb { R } ^ { L _ { p } \times D }$ is a single prompt with token length $L _ { p }$ and the same embedding size $D$ as $\pmb { x } _ { e }$ . Following the notations in Section 3.2, we let $_ { \textbf { \em x } }$ and $\pmb { x } _ { e } = \bar { f } _ { e } ( \pmb { x } )$ be the input and its corresponding method is general enough to the task-agnostic setting. Denoting embedding feature, respectively. Note that we omit the task index $\{ s _ { i } \} _ { i = 1 } ^ { N }$ $t$ of as a subset of $_ { \textbf { \em x } }$ in our notation as our $N$ indices from $[ 1 , M ]$ , we can then adapt the input embedding as follows:
+
+$$
+\pmb { x } _ { p } = [ P _ { s _ { 1 } } ; \cdot \cdot \cdot ; P _ { s _ { N } } ; \pmb { x } _ { e } ] , \quad 1 \leq N \leq M ,
+$$
+
+where ; represents concatenation along the token length dimension. $P$ are free to compose, so they can jointly encode knowledge (e.g. visual features or tasks) for the model to process. Ideally, we
+
+want to achieve a more fine-grained knowledge sharing scheme via prompt combinations at the instance-wise level: similar inputs tend to share more common prompts, and vice versa. We next elaborate our prompt selection strategy and training in the following sections.
+
+# 4.2 INSTANCE-WISE PROMPT QUERY
+
+We design a key-value pair based query strategy to dynamically select suitable prompts for different inputs. This key-valued memory query mechanism shares some design principles with methods in other fields, such as Differentiable Neural Computer (Graves et al., 2016) and VQ-VAE (Oord et al., 2017), which have external memory to maintain, and employs them for a different purpose. With a slight abuse of notation, we associate each prompt as value to a learnable key: $\mathbf { \bar { P } } \overset { ^ { - } } { = } \{ ( k _ { 1 } , P _ { 1 } ) , ( k _ { 2 } , P _ { 2 } ) , \cdots , ( k _ { M } , P _ { M } ) \}$ , where $\pmb { k } \in \mathbb { R } ^ { D _ { k } }$ . Ideally, we would like to let the input instance itself decide which prompts to choose through query-key matching. To this end, we introduce a query function $q : \mathbb { R } ^ { \hat { H } \times W \times \hat { C } }  \mathbb { R } ^ { D _ { k } }$ that encodes input $_ { \textbf { \em x } }$ to the same dimension as the key. Moreover, $q$ should be a deterministic function with respect to different tasks and has no learnable parameters. We directly use the whole pretrained model as a frozen feature extractor to get the query features: $q ( { \pmb x } ) = f ( { \pmb x } ) [ 0 , : ]$ (we use the feature vector corresponding to [class]). Other feature extractors like ConvNet are feasible.
+
+Denote $\gamma : \mathbb { R } ^ { D _ { k } } \times \mathbb { R } ^ { D _ { k } } \to \mathbb { R }$ as a function to score the match between the query and prompt key (we find cosine distance works well). Given an input $_ { \textbf { \em x } }$ , we use $q ( { \pmb x } )$ to lookup the top- $N$ keys by simply solving the objective:
+
+$$
+{ \bf P } _ { { \pmb x } } = \underset { \{ s _ { i } \} _ { i = 1 } ^ { N } \subseteq [ 1 , M ] } { \arg \operatorname* { m i n } } \quad \sum _ { i = 1 } ^ { N } \gamma \left( q ( { \pmb x } ) , { \pmb k } _ { { s } _ { i } } \right) .
+$$
+
+Note that the design of this key-value strategy decouples the query mechanism learning and prompt learning processes, which has been experimentally shown to be critical (see Section 5.3). Furthermore, querying prompts is done in an instance-wise fashion, which makes the whole framework task-agnostic, meaning that the method works without needing clear task boundaries during training, nor task identifications at test time.
+
+Optionally diversifying prompt-selection. Although our method does not need task boundary information, in real-world scenarios and experimental datasets, it is quite common that the task transition is discrete and so task boundaries are known at train time. We find that adding such a prior into our framework can help the model learn better task-specific prompts, especially when tasks have high diversity. To this end, we propose an additional technique for adding task boundaries which is optional for the L2P framework.
+
+During training of task $t$ , we maintain a prompt frequency table $H _ { t } = [ h _ { 1 } , h _ { 2 } , \cdot \cdot \cdot , h _ { M } ]$ , where each entry represents the normalized frequency of prompt $P _ { i }$ being selected up until task $t - 1$ . To encourage the query mechanism select diverse prompts, we modify equation 3 to
+
+$$
+\mathbf { P } _ { \pmb { x } } = \operatorname * { a r g m i n } _ { \{ s _ { i } \} _ { i = 1 } ^ { N } \subseteq [ 1 , M ] } \quad \sum _ { i = 1 } ^ { N } \gamma \left( q ( \pmb { x } ) , \pmb { k } _ { s _ { i } } \right) \cdot h _ { s _ { i } } ,
+$$
+
+where $h _ { s _ { i } }$ penalizes the frequently-used prompts being selected to encourage diversified selection.   
+Equation 4 is only applicable during training; at test time, only equation 3 is needed.
+
+# 4.3 OPTIMIZATION OBJECTIVE FOR L2P
+
+At every training step, after selecting $N$ prompts following the aforementioned query strategy, the adapted embedding feature $\mathbf { \boldsymbol { x } } _ { p }$ is fed into the rest of the pretrained model $f _ { r }$ and the final classifier $g _ { \phi }$ parametrized by $\phi$ . Overall, we seek to minimize the end-to-end training loss function:
+
+$$
+\operatorname* { m i n } _ { \mathbf { P } , \phi } \quad \mathcal { L } \big ( g _ { \phi } \big ( f _ { r } ^ { \mathrm { a v g } } ( x _ { p } ) \big ) , y \big ) + \lambda \sum _ { \mathbf { P } _ { x } } \gamma \left( q ( x ) , k _ { s _ { i } } \right) , \quad s . t . , ~ \mathbf { P } _ { x } \mathrm { ~ i s ~ o b t a i n e d ~ w i t h ~ e q u a t i o n } \ 3 ,
+$$
+
+where $f _ { r } ^ { \mathrm { a v g } } = \mathrm { A v g P o o l } ( f _ { r } ( \pmb { x } _ { p } ) [ N \cdot L _ { p } , : ] )$ , i.e., the output hidden vectors corresponding to the $N \cdot L _ { p }$ prompt locations are averaged before the classification head. The first term is the softmax cross-entropy loss, the second term is a surrogate loss to pull selected keys closer to corresponding query features. $\lambda$ is a scalar to weight the loss.
+
+# 5 EXPERIMENTS
+
+To evaluate the proposed L2P, we closely follow the settings proposed in prior works (Lopez-Paz & Ranzato, 2017; Zeno et al., 2018; Van de Ven & Tolias, 2019), and conduct comprehensive experiments. In particular, we consider (1) the class-incremental setting, where the task identity is unknown during inference; (2) the domain-incremental setting, where the input domain shifts over time; (3) the task-agnostic setting, where there is no clear task boundary. Moreover, we conduct extensive ablation studies to provide a deeper understanding of our method.
+
+Evaluation metrics. For settings with task boundaries and where each task has an associated test set, we use two metrics, Average accuracy $( A )$ and Forgetting $( F )$ , which are widely used in previous works (Lopez-Paz & Ranzato, 2017; Chaudhry et al., 2018; Mai et al., 2021). Denoting by $\mathbf { \Psi } _ { a _ { t , i } }$ the accuracy of the $i$ -th task after finishing training on task $t$ , we can compute the corresponding average accuracy $A _ { t }$ and forgetting $F _ { t }$ up until the current task $t$ as follows:
+
+$$
+A _ { t } = \frac { 1 } { t } \sum _ { i = 1 } ^ { t } a _ { t , i } , \quad F _ { t }  &  = \frac { 1 } { t - 1 } \sum _ { i = 1 } ^ { t - 1 } \operatorname* { m a x } _ { i ^ { \prime } \in \{ 1 , \cdots , t - 1 \} } \left( a _ { i ^ { \prime } , i } - a _ { t , i } \right) .
+$$
+
+We report the final performance $A _ { T }$ and $F _ { T }$ after training on all $T$ tasks. For settings without task boundary or where there is only a single test set available, we only report the final test accuracy following the protocol in previous work (Lomonaco & Maltoni, 2017; Shanahan et al., 2021).
+
+Comparing methods. We compare L2P against several baselines and state-of-the-art continual learning methods. Note that we used the same pretrained ViT-B/16 model (Dosovitskiy et al., 2021) as a starting point for every method to ensure fair comparison. (1) FT-iid is the usual supervised finetuning under the i.i.d. setting, which is the possible upper bound performance a continual learning method could achieve. (2) FT-seq-frozen is the naive sequential fine-tuning approach with the pretrained model frozen. (3) FT-seq is the naive sequential fine-tuning approach (model weights are updated). (4) EWC (Kirkpatrick et al., 2017) is a regularization-based approach aiming at limiting the learning rate of parameters that are important for previous tasks. (5) LwF (Li & Hoiem, 2017) applies the idea of knowledge distillation (Hinton et al., 2015) to preserve knowledge from past tasks. To further demonstrate the effectiveness of our method, we introduce two state-of-the-art rehearsal-based methods, which require additional memory buffer to save samples from past tasks: (6) ER (Chaudhry et al., 2019; Hayes et al., 2019) mixes samples from buffer with samples the from current task in the training process. (7) GDumb (Prabhu et al., 2020) simply constructs the buffer from the sequence of tasks and trains on the buffered samples jointly, so forgetting metric is not applicable to this method. GDumb can outperform many state-of-the-art methods under various settings (Prabhu et al., 2020; Mai et al., 2021). Following the experiment setting in Prabhu et al. (2020), we store an average of 50 samples per class, e.g., a buffer size of 5,000 for CIFAR100, as this is a relatively large choice of buffer size that guarantees SOTA performance.
+
+Experiment details. For L2P, we train all models using Adam (Kingma & Ba, 2014) with $\beta _ { 1 } = 0 . 9$ and $\beta _ { 2 } ~ = ~ 0 . 9 9 9$ , a batch size of 128, and a constant learning rate of 0.03 for all settings. Input images are resized to $2 2 4 \times 2 2 4$ and normalized to the range of $[ 0 , 1 ]$ to match the pretraining setting. As pointed out by Buzzega et al. (2020), training multiple epochs for each task disentangles the effects of possible underfitting from forgetting. Thus, we train every task for 5 epochs in the class- and domain-incremental settings. However, in the task-agnostic setting where we don’t have the concept of a task, we follow Shanahan et al. (2021) to train every batch only once. We set $M = 1 0 ^ { - } N = 5 , L _ { p } = 5$ for all CIFAR-100 based datasets and CORe50. For 5-datasets, we use $M = 2 0 , N = 4 , L _ { p } = 5$ . Prompts only add 46, 080 and 92, 160 parameters to the original pretrained model for these two settings, leading to a small $0 . 0 5 \%$ and $0 . 1 1 \%$ total parameter increase, respectively. We find $\lambda$ in equation 5 is not sensitive and works well in a large range, so we set $\lambda = 0 . 5$ consistently for all datasets.
+
+# 5.1 RESULTS ON CLASS-INCREMENTAL LEARNING
+
+Split CIFAR-100. This dataset randomly splits the original CIFAR-100 dataset (Krizhevsky et al., 2009) into 10 tasks, where each task consist of 10 disjoint classes. Since the tasks are from a single original dataset, they share some similarities and some classes are even from the same superclass.
+
+5-datasets. This dataset (Ebrahimi et al., 2020) consists of five image classification datasets: CIFAR-10, MNIST (LeCun, 1998), Fashion-MNIST (Xiao et al., 2017), SVHN (Netzer et al., 2011), and notMNIST (Bulatov, 2011). Although each dataset alone is not hard, the sequential training of them is fairly challenging to even ImageNet pre-trained models, since models are more susceptible to forgetting when the tasks are diverse (Mehta et al., 2021). We apply the optional strategy introduced in 4.2 to enhance prompt selection diversity.
+
+Table 1: Results on class-incremental learning. Accuracy and forgetting are reported. All methods start from the same pre-trained ViTB/16 model and train on each task for 5 epochs. Methods are separated based on whether rehearsal is applied. All results are shown in percentage $( \% )$ and are averaged over 3 runs.   
+
+<table><tr><td rowspan="2">Method</td><td colspan="2">Split CIFAR-100</td><td colspan="2"> 5-datasets</td></tr><tr><td>Average Acc (↑) </td><td>Forgetting (↓)</td><td>Average Acc (↑)</td><td>Forgetting (↓)</td></tr><tr><td colspan="5">Upper bound:</td></tr><tr><td>FT-iid</td><td>90.85±0.12</td><td></td><td>93.93±0.18</td><td></td></tr><tr><td colspan="5">Non-rehearsal based methods:</td></tr><tr><td>FT-seq-frozen</td><td>17.72±0.34</td><td>59.09±0.25</td><td>39.49±0.12</td><td>42.62±0.20</td></tr><tr><td>FT-seq</td><td>33.61±0.85</td><td>86.87±0.20</td><td>20.12±0.42</td><td>94.63±0.68</td></tr><tr><td>EWC LwF</td><td>47.01±0.29</td><td>33.27±1.17</td><td>50.93±0.09</td><td>34.94±0.07</td></tr><tr><td>L2P (ours)</td><td>60.69±0.63</td><td>27.77±2.17</td><td>47.91±0.33</td><td>38.01±0.28</td></tr><tr><td></td><td>83.83±0.04</td><td>7.63±0.30</td><td>81.14 ±0.93</td><td>4.64 ±0.52</td></tr><tr><td colspan="5">Rehearsal based methods:</td></tr><tr><td>ER</td><td>82.53±0.17</td><td>16.46±0.25</td><td>89.30±0.94</td><td>8.08±0.53</td></tr><tr><td>GDumb</td><td>81.67±0.02</td><td>-</td><td>70.76±0.12</td><td>1</td></tr><tr><td>L2P-R (ours)</td><td>86.31±0.59</td><td>5.83±0.61</td><td>91.92±0.78</td><td>3.34±0.71</td></tr></table>
+
+Table 2: Results on task-agnostic continual learning, in terms of test accuracy. We use Gaussian scheduled CIFAR-100 as the evaluation benchmark. All results are shown in percentage $( \% )$ and are averaged across 3 runs.   
+
+<table><tr><td>Category</td><td>Method</td><td>Test Acc (↑)</td></tr><tr><td>Upper bound</td><td>FT-iid</td><td>90.85±0.12</td></tr><tr><td rowspan="2">Rehearsal</td><td>ER</td><td>82.53±0.17</td></tr><tr><td>GDumb</td><td>81.67±0.02</td></tr><tr><td rowspan="3">Non-rehearsal</td><td>EWC</td><td>63.04±0.42</td></tr><tr><td>LwF</td><td>69.46±0.35</td></tr><tr><td>L2P (ours)</td><td>88.34±0.14</td></tr></table>
+
+Table 3: Results on domain-incremental learning, in terms of test accuracy. We use CORe50 as the evaluation benchmark. All results are shown in percentage $( \% )$ and are averaged across 3 runs.   
+
+<table><tr><td>Category</td><td>Method</td><td>Test Acc (↑)</td></tr><tr><td>Upper bound</td><td>FT-iid</td><td>82.15 ±0.37</td></tr><tr><td rowspan="2">Rehearsal</td><td>ER</td><td>80.10±0.56</td></tr><tr><td>GDumb</td><td>74.92±0.25</td></tr><tr><td rowspan="3">Non-rehearsal</td><td>EWC</td><td>74.82±0.60</td></tr><tr><td>LwF</td><td>75.45±0.40</td></tr><tr><td>L2P (ours)</td><td>78.33±0.06</td></tr></table>
+
+Table 1 summarizes the results on these two class-incremental benchmarks. Similar to what Mehta et al. (2021) have shown: in the simpler task-incremental setting, pre-trained models can overall improve these benchmarks when integrated with existing methods. However, the forgetting rate remains prominent in the class-incremental setting as we shown, suggesting the importance of innovating technologies in pre-trained models beyond applying existing methods.
+
+Our method, L2P, achieves superior performance in terms of both average accuracy and forgetting. In particular, our method: (1) outperforms all non-rehearsal based methods by a large margin, including beating rehearsal-based methods on split CIFAR-100 without rehearsal; and (2) our method improves upon state-of-the-art rehearsal-based methods when incorporating the rehearsal strategy, closing a significant part of the gap to the upper bound performance when doing finetuning under the i.i.d. setting; and (3) compared to the performance of FT-seq-frozen with our method, we can see that naive sequential training is not able to fully take advantage of the pretrained features, further demonstrating the advantages of introducing the prompting strategy.
+
+Table 4: Ablation study on 5-datasets. All results are shown in percentage $( \% )$ .   
+
+<table><tr><td>Method</td><td colspan="2">5-datasets</td></tr><tr><td></td><td>Average Acc (↑)</td><td>Forgetting (↓)</td></tr><tr><td>L2P without prompt pool</td><td>51.96</td><td>26.60</td></tr><tr><td>L2P without key-value pair</td><td>58.33</td><td>20.45</td></tr><tr><td>L2P without diversified prompt selection</td><td>62.26</td><td>17.84</td></tr><tr><td>L2P</td><td>81.14</td><td>4.64</td></tr></table>
+
+![](images/a5b18375003ad8be65525bc772163ea949f3a3366ea66a4212b70750bd154e79.jpg)  
+Figure 3: Prompt selection histograms for (left) Split CIFAR-100 and (right) 5-datasets. Note that we only show the first 5 tasks for Split CIFAR-100 for better readability.
+
+# 5.2 RESULTS ON TASK-AGNOSTIC AND DOMAIN INCREMENTAL SETTINGS
+
+Gaussian scheduled CIFAR-100. In this task-agnostic setting, the distribution of data shifts gradually throughout the learning process (Shanahan et al., 2021), the probability that a class is present in a batch follows a Gaussian distribution centered at some time step. There is no explicit task boundaries between batches, thus requiring methods to be able to implicitly adapt to non-stationary data distribution without utilizing any task-specific information during training and inference.
+
+Table 2 summarizes the results. L2P achieves the best performance among all methods, including rehearsal based ones. The task-agnostic setting is usually considered more challenging than the class-incremental setting. Since these two benchmarks have the same test test, we can compare them deeper. Interestingly, EWC and LwF both achieve higher accuracy than that on split CIFAR-100, indicating that a well-pretrained model itself may serve as a better starting point for task-agnostic continual learning. Similar observations has been reported on a simpler task-incremental setting in Mehta et al. (2021). Moreover, L2P achieves a test accuracy $8 8 . 3 4 \%$ , which is very close to the upper bound performance $9 0 . 8 5 \%$ shown in Table 1, suggesting strongly reduced forgetting rate.
+
+CORe50. This is a dataset specifically designed for continual object recognition (Lomonaco & Maltoni, 2017). It is a collection of 50 objects collected in 11 distinct domains, where 8 of them (120,000 samples) are used for training, and the rest are considered as a single test set (45,000 examples). Methods are trained on each domain sequentially.
+
+Table 3 summarizes the results on the domain-incremental setting. Although L2P still achieves better performance than most methods, surprisingly, all methods are quite close to the upper bound performance FT-iid. This indicates that a well pretrained model has the potential to accumulate knowledge from different domains without much interference. However, more comprehensive experiments are required to further confirm this observation, which we leave to future work.
+
+# 5.3 EFFECTIVENESS OF CORE DESIGNS
+
+We further conduct ablation studies to demonstrate the effectiveness of the core designs of L2P.
+
+Prompt pool. To further confirm the importance of the prompt pool, we design a counterpart of our method with only a single prompt instead of the prompt pool. This variation of our method keeps the same prompt capacity as L2P in equation 2. From Table 4 (row 1 and 4), we can see that L2P significantly outperforms its counterpart with a single prompt, suggesting that the prompt pool encodes task-relevant and task-specific knowledge well.
+
+![](images/791ba5f7100cc72c11bc599dc9c7f189726cce930a5b6b7ffeabcddbd4201067.jpg)  
+Figure 4: Left-Middle: Average accuracy w.r.t prompt length $L _ { p }$ and prompt selection size $N$ for Split CIFAR-100 and 5-datasets, respectively, given $M = 2 0$ . Right: Average accuracy $( \% )$ w.r.t. prompt pool size $M$ , given $L _ { p } = 5$ , $N = 5$ for Split CIFAR-100 and $L _ { p } = 5$ , $N = 4$ for 5-datasets.
+
+Key-value pair design. We remove the learnable key associated with prompts and directly use mean of prompts as keys and the mean of input embedding as query features, as they reside in the same space. From Table 4 (row 2), we can see this results in a significant drop, demonstrating the importance of introducing learnable keys to decouple the query and prompt learning process.
+
+Diversified prompt selection. This technique is used by default on 5-dataset only. When we remove it, (Table 4 row 3), we basically allow instances from different tasks to choose prompts freely. The decrease in performance demonstrates that when tasks are diverse, adding the diversified prompt selection strategy can indeed reduce unnecessary knowledge sharing and thus mitigating interference between unrelated tasks.
+
+To better understand the prompt selection mechanism, we plot the prompt selection histograms for each task in both split CIFAR-100 and 5-datasets in Figure 3 under the best-performing parameters settings, respectively. From the plot of Split CIFAR-100 (left), the tasks largely share all prompts, meaning that our prompt selection mechanism encourages more knowledge sharing between similar tasks. In contrast, in the plot of 5-datasets (right), diverse tasks tends to choose more task-specific prompts and share less.
+
+Effect of hyperparameters for L2P. Recall that there are three key hyperparameters, including the size of the prompt pool $M$ , length of a single prompt $L _ { p }$ , and the selection size $N$ used as model input. Intuitively, $M$ decides the total capacity of learnable prompt parameters. $L _ { p }$ decides capacity of a singe prompt (which jointly encodes certain knowledge), and $L _ { p } \times N$ decides the total size used to prepend the input. From the results on both datasets (Figure 4 (left-middle)), a smaller $L _ { p }$ always negatively affects results. We hypothesize that a reasonable capacity of a single prompt is critical to encode a certain aspect of shared knowledge. Increasing the prompt pool size shows positive effect for performance as shown in Figure 4 (right), especially on 5-datasets, suggesting a large enough pool size is needed to encode task-specific knowledge when tasks are diverse.
+
+# 6 CONCLUSION
+
+This paper presents a novel method to address some of the key challenges in continual learning with a method that can achieve strong performance without a need for rehearsal and task identity. L2P introduces prompt-based learning to continual learning and proposes a novel technique to enable a single pre-trained model to adapt to sequential tasks via a shared prompt pool, successfully mitigating the catastrophic forgetting problem. The resulting method achieves good results on challenging continual learning problems, including class-incremental, domain-incremental, and task-agnostic settings, demonstrating the effectiveness of the method, as well as its advantages to satisfy the practical data privacy requirement when storing data as rehearsal buffer is prohibited.
+
+Although our method is demonstrated on vision models, it does not make any assumption of modalities. We leave exploration on other modalities as future work. Additionally, L2P assumes there are pre-trained sequence-based models. While they have become common assets in advanced communities, how to generalize our framework to ConvNets could another appealing research direction.
+
+# 7 ETHICS STATEMENT
+
+L2P is a strong continual learning method and has great potential to be applied in various fields. However, there are some ways it could be misused. Our method takes a well-pretrained model as a backbone, thus any bias and fairness issues (Mehrabi et al., 2021) in the original model may be carried over during the continual learning process. We encourage any users to thoroughly check the pretrained model to mitigate any bias and fairness issues. Moreover, the method could be deployed in safety-critical applications, such as autonomous driving systems (Grigorescu et al., 2020), which may present potential security issues in terms of adversarial attacks (Madry et al., 2017). We would recommend testing the robustness of our method in future work and design corresponding defense techniques to deal with potential security concerns.
+
+# 8 REPRODUCIBILITY
+
+To make the results presented in our work reproducible, we include all experiment setups and details, evaluation metrics, and comparing methods in Section 5. We test our method on multiple publicly available datasets and under different settings. We report the average and corresponding standard deviations over multiple runs using different randoms seeds for our main results (Table 1, 2 and 3). Our results are also verified on different hardwares, including TPU and GPU. We plan to make the code publicly available upon acceptance.
+
+# REFERENCES
+
+Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuytelaars. Memory aware synapses: Learning what (not) to forget. In ECCV, 2018. 2
+
+Yaroslav Bulatov. notmnist dataset, 2011. URL http://yaroslavvb.blogspot.com/ 2011/09/notmnist-dataset.html. 7
+
+Pietro Buzzega, Matteo Boschini, Angelo Porrello, Davide Abati, and Simone Calderara. Dark experience for general continual learning: a strong, simple baseline. In NeurIPS, 2020. 6
+
+Arslan Chaudhry, Marc’Aurelio Ranzato, Marcus Rohrbach, and Mohamed Elhoseiny. Efficient lifelong learning with a-gem. arXiv preprint arXiv:1812.00420, 2018. 3, 6
+
+Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny, Thalaiyasingam Ajanthan, Puneet K Dokania, Philip HS Torr, and Marc’Aurelio Ranzato. On tiny episodic memories in continual learning. arXiv preprint arXiv:1902.10486, 2019. 3, 6
+
+Matthias Delange, Rahaf Aljundi, Marc Masana, Sarah Parisot, Xu Jia, Ales Leonardis, Greg Slabaugh, and Tinne Tuytelaars. A continual learning survey: Defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021. 2
+
+Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. ICLR, 2021. 3, 6
+
+Sayna Ebrahimi, Franziska Meier, Roberto Calandra, Trevor Darrell, and Marcus Rohrbach. Adversarial continual learning. In ECCV, 2020. 6
+
+Alex Graves, Greg Wayne, Malcolm Reynolds, Tim Harley, Ivo Danihelka, Agnieszka GrabskaBarwinska, Sergio G ´ omez Colmenarejo, Edward Grefenstette, Tiago Ramalho, John Agapiou, ´ et al. Hybrid computing using a neural network with dynamic external memory. Nature, 538 (7626):471–476, 2016. 5
+
+Sorin Grigorescu, Bogdan Trasnea, Tiberiu Cocias, and Gigel Macesanu. A survey of deep learning techniques for autonomous driving. Journal of Field Robotics, 37(3):362–386, 2020. 10
+
+Raia Hadsell, Dushyant Rao, Andrei A Rusu, and Razvan Pascanu. Embracing change: Continual learning in deep neural networks. Trends in cognitive sciences, 2020. 1, 4
+
+Tyler L Hayes, Nathan D Cahill, and Christopher Kanan. Memory efficient experience replay for streaming learning. In ICRA, 2019. 3, 6
+
+Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015. 6
+
+Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014. 6
+
+James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks. PNAS, 114(13):3521–3526, 2017. 2, 6
+
+Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009. 6
+
+Ankit Kumar, Ozan Irsoy, Peter Ondruska, Mohit Iyyer, James Bradbury, Ishaan Gulrajani, Victor Zhong, Romain Paulus, and Richard Socher. Ask me anything: Dynamic memory networks for natural language processing. In ICML, 2016. 3
+
+Yann LeCun. The mnist database of handwritten digits. http://yann. lecun. com/exdb/mnist/, 1998. 6
+
+Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning. arXiv preprint arXiv:2104.08691, 2021. 1, 3, 4
+
+Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021. 1, 3
+
+Zhizhong Li and Derek Hoiem. Learning without forgetting. TPAMI, 40(12):2935–2947, 2017. 2, 6
+
+Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. Pretrain, prompt, and predict: A systematic survey of prompting methods in natural language processing. arXiv preprint arXiv:2107.13586, 2021. 1, 3, 4
+
+Vincenzo Lomonaco and Davide Maltoni. Core50: a new dataset and benchmark for continuous object recognition. In Conference on Robot Learning, 2017. 6, 8
+
+David Lopez-Paz and Marc’Aurelio Ranzato. Gradient episodic memory for continual learning. NeurIPS, 2017. 6
+
+Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks. arXiv preprint arXiv:1706.06083, 2017. 10
+
+Zheda Mai, Ruiwen Li, Jihwan Jeong, David Quispe, Hyunwoo Kim, and Scott Sanner. Online continual learning in image classification: An empirical survey. arXiv preprint arXiv:2101.10423, 2021. 1, 3, 6
+
+Arun Mallya and Svetlana Lazebnik. Packnet: Adding multiple tasks to a single network by iterative pruning. In CVPR, 2018. 3
+
+Bryan McCann, Nitish Shirish Keskar, Caiming Xiong, and Richard Socher. The natural language decathlon: Multitask learning as question answering. arXiv preprint arXiv:1806.08730, 2018. 3
+
+Michael McCloskey and Neal J Cohen. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, pp. 109–165. Elsevier, 1989. 1
+
+Ninareh Mehrabi, Fred Morstatter, Nripsuta Saxena, Kristina Lerman, and Aram Galstyan. A survey on bias and fairness in machine learning. ACM Computing Surveys (CSUR), 54(6):1–35, 2021. 10
+
+Sanket Vaibhav Mehta, Darshan Patil, Sarath Chandar, and Emma Strubell. An empirical investigation of the role of pre-training in lifelong learning. ICML Workshop on Theory and Foundation of Continual Learning, 2021. 7, 8
+
+Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, and Andrew Y Ng. Reading digits in natural images with unsupervised feature learning. In NIPS, 2011. 6
+
+Aaron van den Oord, Oriol Vinyals, and Koray Kavukcuoglu. Neural discrete representation learning. arXiv preprint arXiv:1711.00937, 2017. 5
+
+German I Parisi, Ronald Kemker, Jose L Part, Christopher Kanan, and Stefan Wermter. Continual lifelong learning with neural networks: A review. Neural Networks, 113:54–71, 2019. 1, 3
+
+Ameya Prabhu, Philip HS Torr, and Puneet K Dokania. Gdumb: A simple approach that questions our progress in continual learning. In ECCV, 2020. 6
+
+Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9, 2019. 3
+
+Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. JMLR, 21:1–67, 2020. 1, 3
+
+Andrei A Rusu, Neil C Rabinowitz, Guillaume Desjardins, Hubert Soyer, James Kirkpatrick, Koray Kavukcuoglu, Razvan Pascanu, and Raia Hadsell. Progressive neural networks. arXiv preprint arXiv:1606.04671, 2016. 3
+
+Timo Schick and Hinrich Schutze. Exploiting cloze questions for few shot text classification and ¨ natural language inference. arXiv preprint arXiv:2001.07676, 2020. 3
+
+Murray Shanahan, Christos Kaplanis, and Jovana Mitrovic. Encoders and ensembles for task-free ´ continual learning. arXiv preprint arXiv:2105.13327, 2021. 6, 8
+
+Taylor Shin, Yasaman Razeghi, Robert L Logan IV, Eric Wallace, and Sameer Singh. Autoprompt: Eliciting knowledge from language models with automatically generated prompts. arXiv preprint arXiv:2010.15980, 2020. 1
+
+Reza Shokri and Vitaly Shmatikov. Privacy-preserving deep learning. In Proc SIGSAC conference on computer and communications security, 2015. 3
+
+Gido M Van de Ven and Andreas S Tolias. Three scenarios for continual learning. arXiv preprint arXiv:1904.07734, 2019. 6
+
+Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. NeurIPS, 2017. 3
+
+Zifeng Wang, Tong Jian, Kaushik Chowdhury, Yanzhi Wang, Jennifer Dy, and Stratis Ioannidis. Learn-prune-share for lifelong learning. In ICDM, 2020. 3
+
+Han Xiao, Kashif Rasul, and Roland Vollgraf. Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. arXiv preprint arXiv:1708.07747, 2017. 6
+
+Jaehong Yoon, Eunho Yang, Jeongtae Lee, and Sung Ju Hwang. Lifelong learning with dynamically expandable networks. arXiv preprint arXiv:1708.01547, 2017. 3
+
+Friedemann Zenke, Ben Poole, and Surya Ganguli. Continual learning through synaptic intelligence. In ICML, 2017. 2
+
+Chen Zeno, Itay Golan, Elad Hoffer, and Daniel Soudry. Task agnostic continual learning using online variational bayes. arXiv preprint arXiv:1803.10123, 2018. 6

parse/dev/Vx1JadlOIt/Vx1JadlOIt_content_list.json

@@ -0,0 +1,1269 @@
+[
+    {
+        "type": "text",
+        "text": "VisionLLM: Large Language Model is also an Open-Ended Decoder for Vision-Centric Tasks ",
+        "text_level": 1,
+        "bbox": [
+            197,
+            122,
+            800,
+            172
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "Wenhai Wang∗2 Zhe Chen∗1,3 Xiaokang Chen∗1,4 Jiannan $\\mathbf { W _ { u } } ^ { * 1 , 5 }$ Xizhou Zhu1,6 Gang Zeng4 Ping Luo5 Tong $\\mathbf { L u ^ { 3 } }$ Jie Zhou6 Yu Qiao1 Jifeng Dai†1,6 1OpenGVLab, Shanghai AI Laboratory 2The Chinese University of Hong Kong 3Nanjing University 4Peking University 5The University of HongKong 6Tsinghua University ",
+        "bbox": [
+            186,
+            224,
+            816,
+            285
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "Code: https://github.com/OpenGVLab/VisionLLM ",
+        "bbox": [
+            315,
+            297,
+            686,
+            310
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "Abstract ",
+        "text_level": 1,
+        "bbox": [
+            462,
+            347,
+            535,
+            363
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "Large language models (LLMs) have notably accelerated progress towards artificial general intelligence (AGI), with their impressive zero-shot capacity for user-tailored tasks, endowing them with immense potential across a range of applications. However, in the field of computer vision, despite the availability of numerous powerful vision foundation models (VFMs), they are still restricted to tasks in a pre-defined form, struggling to match the open-ended task capabilities of LLMs. In this work, we present an LLM-based framework for vision-centric tasks, termed VisionLLM. This framework provides a unified perspective for vision and language tasks by treating images as a foreign language and aligning vision-centric tasks with language tasks that can be flexibly defined and managed using language instructions. An LLM-based decoder can then make appropriate predictions based on these instructions for open-ended tasks. Extensive experiments show that the proposed VisionLLM can achieve different levels of task customization through language instructions, from fine-grained object-level to coarse-grained task-level customization, all with good results. It’s noteworthy that, with a generalist LLMbased framework, our model can achieve over $60 \\%$ mAP on COCO, on par with detection-specific models. We hope this model can set a new baseline for generalist vision and language models. The code shall be released. ",
+        "bbox": [
+            233,
+            380,
+            766,
+            627
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "1 Introduction ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            652,
+            310,
+            670
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "The emergence of large language models (LLMs) like ChatGPT [35] has revolutionized the landscape of artificial general intelligence (AGI), showcasing their impressive zero-shot capabilities in addressing various natural language processing (NLP) tasks through user-tailored prompts or language instructions. Despite these advancements, it’s essential to note that the triumph of LLMs does not effortlessly extend to pure vision and vision-language tasks, due to the inherent disparities between modalities and task formats. ",
+        "bbox": [
+            174,
+            678,
+            825,
+            760
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "text",
+        "text": "The field of computer vision presents a unique set of challenges and paradigms that differ from those of NLP. The traditional paradigm of vision foundation models is pre-training followed by fine-tuning [51, 11, 43, 53, 17, 44], which is effective but comes with significant marginal costs when adapting to diverse downstream scenarios. As shown in Figure 1a, while approaches such as multi-task unification [38, 50, 1, 49, 72] have been used to achieve generalist capability, they often struggle to overcome the limitations imposed by pre-defined tasks, resulting in a gap in open-ended task capabilities compared to LLMs. Recently, visual prompt tuning [24, 66, 70, 67, 54] has emerged as a way to flexibly outline some pure vision tasks (see Figure 1b), such as object detection, instance ",
+        "bbox": [
+            174,
+            767,
+            825,
+            877
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "image",
+        "img_path": "images/2a51129098861a5218a07b95eae7f1b82a9343d82a0e8b59c93c816806fb2d8d.jpg",
+        "image_caption": [],
+        "image_footnote": [],
+        "bbox": [
+            181,
+            90,
+            366,
+            231
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "image",
+        "img_path": "images/13482ae826748fd88984f398aaf673054cdcde523ff2f0db87c13f63a13b2711.jpg",
+        "image_caption": [
+            "(b) Visual prompt tuning [24, 56, 54] are inconsistent with the format of LLMs. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            403,
+            90,
+            593,
+            229
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "(a) Vision generalist models [51, 53, 74] are constrained by the format of pre-defined tasks. ",
+        "bbox": [
+            174,
+            236,
+            374,
+            273
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "image",
+        "img_path": "images/6d3aa85a408251041c0bb17b17eb26209941e8a4f6d8ea22b2183fb30ee30d4e.jpg",
+        "image_caption": [
+            "(c) VisionLLM (ours) can flexibly manage vision-centric tasks using language instructions like LLMs. ",
+            "Figure 1: Comparison of our VisionLLM with popular paradigms. Unlike current vision generalist models that depend on pre-defined task formats and visual prompt tuning models that are inconsistent with large language models (LLMs), VisionLLM leverages the power of LLMs for open-ended vision tasks by using language instructions. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            632,
+            92,
+            815,
+            228
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "segmentation, and pose estimation, using visual masking. However, the format of visual prompts considerably deviates from that of language instructions, making it challenging to directly apply the reasoning abilities and world knowledge of LLMs to vision tasks. Therefore, there is an urgent need for a unified generalist framework that can seamlessly integrate the strengths of LLMs with the specific requirements of vision-centric tasks. ",
+        "bbox": [
+            174,
+            359,
+            823,
+            429
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "In this work, we present VisionLLM, a novel framework that aligns the definitions of vision-centric tasks with the methodologies of LLMs. Leveraging the reasoning and parsing capacities of LLMs, VisionLLM is designed to empower open-ended task capabilities for vision-centric tasks. Specifically, it comprises three core components: (1) a unified language instruction designed for vision and vision-language tasks, (2) a language-guided image tokenizer, and (3) an LLM-based open-ended task decoder that orchestrates various tasks using language instructions. With this framework, a wide range of vision-centric tasks can be seamlessly integrated, including object detection, instance segmentation, image captioning, and visual grounding. In addition, the framework also facilitates task customization at different levels of granularity, allowing for the customization of target objects, output formats, task descriptions, etc. ",
+        "bbox": [
+            174,
+            435,
+            825,
+            574
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "Compared to current popular API-based applications [60, 65, 42, 32, 28], our model takes a unified, end-to-end approach to integrate VFMs and LLMs, streamlining and enhancing the overall efficiency of the overall process, and leveraging the strengths and data of both VFMs and LLMs within a single, cohesive system. Furthermore, our model surpasses the limitations of generalist vision models pre-trained on pre-defined tasks. VisionLLM can effectively manage vision-centric tasks through language instructions, embodying a flexible and open-ended approach that is not constrained by pre-set tasks. This versatility makes VisionLLM a robust and powerful generalist model for vision and vision-language tasks, opening up new possibilities for the development of unified generalist models that bridge the domains of vision and language. ",
+        "bbox": [
+            174,
+            580,
+            825,
+            705
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "In summary, our main contributions are as follows: ",
+        "bbox": [
+            174,
+            710,
+            508,
+            724
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "(1) We propose VisionLLM, the first framework that leverages the power of LLMs to address visioncentric tasks in an open-ended and customizable manner. By aligning the definitions of vision-centric tasks with LLM methodologies, VisionLLM breaks new ground in enabling the unified modeling of vision and language, opening up possibilities for advancing the field. ",
+        "bbox": [
+            174,
+            731,
+            825,
+            787
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "(2) We overcome many difficulties when porting LLMs to vision-centric tasks, by designing unified language instruction that matches the format of language models and covers various vision-centric tasks including visual perception. Correspondingly, we develop a language-guided image tokenizer and an LLM-based task decoder that can handle open-ended tasks according to the given language instructions based on the LLMs’ reasoning and parsing capabilities. ",
+        "bbox": [
+            174,
+            794,
+            823,
+            863
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "(3) We construct a series of tasks with different granularities to verify the effectiveness of our models, ranging from easy to hard, and from pre-defined to flexible. Through these validations, we demonstrate the remarkable generality of our models, showcasing their ability to handle diverse ",
+        "bbox": [
+            174,
+            869,
+            823,
+            911
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "text",
+        "text": "Human: “Identify the objects in <image> that belong to {‘What is the child eating?’: <c0>, ‘red gamepad': <c1>} and draw a bounding box around each one. The output should be a list of tuples in ",
+        "bbox": [
+            191,
+            99,
+            330,
+            176
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "image",
+        "img_path": "images/bdd6a545fad93f1fe2171eca15920f046c6833f0db480da5b22ac95467046b01.jpg",
+        "image_caption": [],
+        "image_footnote": [],
+        "bbox": [
+            336,
+            98,
+            472,
+            172
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "the format (c, x1, y1, x2, y2), where c is the class label and x1, y1, x2, y2 are the coordinates of the top-left and bottom-right corners of the bounding box. The coordinates should be within the range 1024.\" ",
+        "bbox": [
+            191,
+            179,
+            473,
+            223
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "VisionLLM: \"The bounding boxes are [(<c0>, 226.4, 229.8, 363.1, 347.4), (<c1>, 441.1, 183.5, 538.6, 269.9)].\" ",
+        "bbox": [
+            191,
+            227,
+            472,
+            251
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "Human: \"For each object in <image> that is a member of {'motorbike': $< \\mathsf { c } 0 > \\}$ , output a tuple with the class label and the coordinates of a polygon with 16 points that ",
+        "bbox": [
+            521,
+            94,
+            660,
+            174
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "image",
+        "img_path": "images/654913c5cf06a5a7be1ab27e563f83275228c04250e18421ce001abe0726ff9d.jpg",
+        "image_caption": [],
+        "image_footnote": [],
+        "bbox": [
+            666,
+            94,
+            810,
+            169
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "encloses the object. The coordinates should be within the range 1024. The output format should be (c, x1, y1, x2, y2, ..., x16, y16).\" ",
+        "bbox": [
+            522,
+            175,
+            805,
+            210
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "VisionLLM: \"The detected polygons are $I ( < \\mathsf { c o } > , 1 3 5 . 3 ,$ 95.7, 123.4, 53.4, 84.9, 57.6, 66.8, 60.5, 60.1, 72.3, 34.2, 71.4, ..., 124.9, 119.3)].\" ",
+        "bbox": [
+            521,
+            213,
+            797,
+            250
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "(a) Object Detection with Customized Class Set ",
+        "bbox": [
+            194,
+            262,
+            475,
+            275
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "Human: \"Provide a long description of the <image>.\" ",
+        "bbox": [
+            191,
+            280,
+            333,
+            304
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "image",
+        "img_path": "images/59e4f4bd19e53294b66fcc2461ba5df514fa99c57a11021c98cb5ddd55b50213.jpg",
+        "image_caption": [],
+        "image_footnote": [],
+        "bbox": [
+            336,
+            277,
+            478,
+            344
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "VisionLLM: \"The image features a white dog laying down on a foldable outdoor ",
+        "bbox": [
+            191,
+            309,
+            333,
+            343
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "cooling pad , ",
+        "text_level": 1,
+        "bbox": [
+            192,
+            344,
+            256,
+            354
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "which is placed on top of a wooden table. The dog has its head in the middle of the cooling pad, and a fan is positioned behind it, blowing air onto the pad. The table is outdoors, in a grassy area, and there are two chairs placed around it, one on each side. The scene appears to be set in a sunny weather, with a bright and sunny sky, which contributes to the overall outdoor atmosphere.\" ",
+        "bbox": [
+            191,
+            356,
+            473,
+            448
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "(b) Object Location with Editable Output Format ",
+        "bbox": [
+            519,
+            262,
+            808,
+            275
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "Human: \"With the objects in the <image>, if you feel hot, what will you do? ",
+        "bbox": [
+            522,
+            290,
+            660,
+            325
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "image",
+        "img_path": "images/1675fe85ffcea4c5dc05f23a7f48b5ed1559a53361c537b2734d6cb4325d41a8.jpg",
+        "image_caption": [],
+        "image_footnote": [],
+        "bbox": [
+            668,
+            287,
+            812,
+            371
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "VisionLLM: \"First, You can also try turning off the computer and other electrical appliances in the ",
+        "bbox": [
+            522,
+            330,
+            660,
+            377
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "room to reduce the heat generated by their operation. As an alternative, you can use the desk fan provided in the image to create a cooling effect in the room. You can plug in the fan and adjust its settings to ensure it blows air in a direction that reaches the room’s occupants effectively.\" ",
+        "bbox": [
+            522,
+            378,
+            802,
+            448
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "(c) Image Description with Controllable Text Length (d) Visual Question Answer with Complex Reasoning ",
+        "bbox": [
+            178,
+            460,
+            490,
+            473
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            506,
+            460,
+            818,
+            473
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "Figure 2: Results and visualizations of our VisionLLM. Guided by language instructions, our unified generalist framework showcases its effectiveness on diverse open-ended vision-centric tasks. The text marked with a gray background indicates the customized instructions and the desired outputs. ",
+        "bbox": [
+            174,
+            481,
+            825,
+            523
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "scenarios, including random object categories, random output formats, and random task descriptions, as shown in Figure 2. The successful outcomes of these validations underscore the tremendous potential of our model in harnessing the capabilities of LLMs to control and guide vision-centric tasks. In addition, with a generalist LLM-based framework, our model also yields promising results on various vision-centric tasks. Notably, our generalist model achieves an impressive mAP score of $60 \\%$ on the COCO dataset, surpassing many detection-specific models [73, 6, 20] and approaching the state-of-the-art record. ",
+        "bbox": [
+            174,
+            537,
+            825,
+            633
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "2 Related Work ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            646,
+            321,
+            662
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "2.1 Large Language Model ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            671,
+            377,
+            686
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "Large language models (LLMs) have gained significant attention in the field of natural language processing (NLP) and artificial general intelligence (AGI), due to their impressive capabilities in language generation, in-context learning, world knowledge, and reasoning. The GPT family, including GPT-3 [5], ChatGPT [35], GPT-4 [34], and InstructGPT [36] are most representative works of LLMs. Other LLMs like OPT [69], LLaMA [46], MOSS [14], and GLM [68] have also made substantial contributions to the field. These models achieve high performance and are open-sourced, serving as valuable resources for training large models and as foundations for further fine-tuning for specific purposes. For instance, Alpaca [45] introduces a self-instruct framework that facilitates instruction tuning of the LLaMA model, reducing the reliance on human-written instruction data. Recently, the emergence of these LLMs has also opened up API-based applications for solving vision-centric tasks. These applications have integrated visual APIs with language models to enable decision-making or planning based on visual information, such as Visual ChatGPT [60], MM-REACT [65], HuggingGPT [42], InternGPT [32], and VideoChat [28]. However, despite the convenience of using language-based instructions to define tasks and describe visual elements, these interactive systems [60, 65, 42, 32, 28] still face limitations in capturing fine-grained visual details and understanding complex visual contexts, which hinder their ability to effectively connecting vision and language models. In summary, while LLMs have shown tremendous potential in various NLP applications, their applicability to vision-centric tasks has been limited by the challenges posed by modalities and task formats. ",
+        "bbox": [
+            174,
+            690,
+            825,
+            911
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            174,
+            90,
+            823,
+            132
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "2.2 Vision Generalist Model ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            148,
+            382,
+            164
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "The pursuit of generalist models [74, 33, 62], which aim to handle a wide range of tasks using a shared architecture and parameters, has been a long-standing goal in the machine learning community. Inspired by the success of sequence-to-sequence (seq2seq) models in the field of NLP [38], recent advancements such as OFA [50], Flamingo [1], and GIT [49] propose modeling diverse tasks as sequence generation tasks. Unified-IO [33], Pix2Seq v2 [8], and UniTab [63] extend this idea by using discrete coordinate tokens to encode and decode spatial information for more tasks. Gato [39] also incorporates reinforcement learning tasks into the seq2seq framework, while GPV [19] develops a general-purpose vision system by combining a seq2seq module with a DETR-based visual encoder [6]. However, these methods suffer from some limitations, such as slow inference speed and performance degradation due to the non-parallel auto-regressive decoding process. Uni-Perceivers [74, 72, 26] solve these issues by unifying different tasks using the maximum likelihood target for each input based on representation similarity, regardless of their modality, making it possible to support both generation and non-generation tasks in a unified framework. Nevertheless, these generalist models are still restricted by pre-defined tasks and cannot support flexible open-ended task customization based on language instructions like LLMs. ",
+        "bbox": [
+            174,
+            170,
+            825,
+            377
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "2.3 Instruction Tuning ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            393,
+            344,
+            407
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "Language instructions are a powerful way to express various NLP tasks and examples for LLMs, as introduced by GPT-3 [5]. Following this idea, subsequent works, such as InstructGPT [36], FLAN [13, 59], and OPT-IML [23], explore the instruction-tuning method [58, 57] and demonstrate that this simple approach effectively enhances the zero-shot and few-shot capabilities of LLMs. The language instruction paradigm has also been adopted by the computer vision community to define image-to-text tasks. Flamingo [1] is a milestone work that uses vision and language inputs as prompts and achieves remarkable few-shot results in various vision-language tasks, such as image captioning [9] and VQA [2]. BLIP-2 [27] further connects the visual encoder with LLMs through a querying transformer and a linear projection layer to build strong multimodal models. MiniGPT-4 [71] and LLaVA [30] finetune the BLIP-2-style models on synthetic multimodal instruction-following data to unleash the potential of LLMs. However, these models mainly focus on image-to-text tasks and fail to address visual perception, such as object detection, instance segmentation, pose estimation, etc. To tackle image inpainting tasks, Bar et al. [3] introduces the first visual prompting framework that utilizes inpainting with discrete tokens on images. Painter [55] and SegGPT [56] employ masked image modeling on raw pixels for in-context learning with paired images. While these visual prompt models demonstrate good results in segmentation tasks, their applicability to numerous real-world vision tasks is challenging. Moreover, defining the visual prompts as image inpainting is inconsistent with the language instructions in LLMs, hard to leverage the reasoning, parsing ability, and world knowledge of LLMs. In this work, we aim to align vision-centric tasks with language tasks, use language instructions to unifiedly and flexibly define all tasks, and solve them with a shared LLM-based task decoder. ",
+        "bbox": [
+            174,
+            414,
+            825,
+            704
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "3 VisionLLM ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            723,
+            302,
+            739
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "3.1 Overall Architecture ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            751,
+            357,
+            766
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "This work targets to provide a unified generalist framework that can seamlessly integrate the strengths of large language models (LLMs) with the specific requirements of vision-centric tasks. As shown in Figure 3, the overall architecture of VisionLLM consists of three key designs: (1) a unified language instruction that provides a consistent interface for vision-centric task definition and customization; (2) a language-guided image tokenizer, which encodes visual information in alignment with the given language prompt, enabling the model to comprehend and parse the visual content effectively; and (3) an LLM-based open-task decoder, which utilizes the encoded visual information and language instructions to generate satisfactory predictions or outputs. The three designs work together to achieve a flexible and open-ended framework that can handle various vision-centric tasks at different levels of task customization through language instructions. ",
+        "bbox": [
+            173,
+            772,
+            825,
+            911
+        ],
+        "page_idx": 3
+    },
+    {
+        "type": "text",
+        "text": "Vision-language example: \"Describe the image <image> in details.\" Language Instructions <text> Vision-only example: \"For each object in image <image> that is a member of class set <class>, output a tuple with the class label and the coordinates of a polygon with 16 points that encloses the object. The coordinates should be within range <range>. The output format should be (c, x1, y1, ...).\" ",
+        "bbox": [
+            194,
+            214,
+            800,
+            275
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "image",
+        "img_path": "images/b0780942b0411460c047304f6e85746e7bf2043070c0a7418953fc1b0cc38a4d.jpg",
+        "image_caption": [
+            "Figure 3: Overall architecture of the proposed VisionLLM. It consists of three parts: a unified language instruction designed to accommodate both vision and vision-language tasks, an image tokenizer that encodes visual information guided by language instructions, and an LLM-based openended task decoder that executes diverse tasks defined by language instructions. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            194,
+            93,
+            803,
+            210
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "Different from previous interactive systems [60, 65, 42, 32, 28] that rely on APIs, our VisionLLM presents a more flexible and end-to-end pipeline. Given language instructions that describe the current tasks and an input image, the model first uses a language-guided image tokenizer to encode the image tokens based on the given prompt. Then, the image tokens and language instructions are fed to an LLM-based open-ended task decoder. Finally, it evaluates the generated outputs against the task definition given by the unified language instructions, enabling the model to produce task-specific results. This seamless, end-to-end pipeline enables VisionLLM to effectively combine vision and language, achieving remarkable performance in open-ended and customizable vision-centric tasks. ",
+        "bbox": [
+            174,
+            353,
+            825,
+            464
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "3.2 Unified Language Instruction ",
+        "text_level": 1,
+        "bbox": [
+            176,
+            477,
+            418,
+            492
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "We first introduce unified language instructions to describe vision-centric tasks. This design enables the unification of various vision-only and vision-language task descriptions and allows for flexible task customization. ",
+        "bbox": [
+            174,
+            497,
+            825,
+            539
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "Vision-Language Tasks. The instructions for vision-language tasks such as image captioning and visual question answering (VQA) are straightforward and similar to NLP tasks. Following previous methods [27, 74, 30], we describe the image captioning task like “The image is <image>. Please generate a caption for the image: ”, and the VQA task like “The image is <image>. Please generate an answer for the image according to the question: <question>”. Here, <image> and <question> are the placeholders of the image tokens and the question, respectively. The image tokens are directly placed at the placeholder <image>. ",
+        "bbox": [
+            174,
+            545,
+            825,
+            643
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "Vision-Only Tasks. Designing effective language instructions for vision tasks is a challenging endeavor due to the differences in modality and task format between vision and language. Here, we describe vision tasks by providing a task description and specifying the desired output format via language instructions. ",
+        "bbox": [
+            173,
+            648,
+            825,
+            704
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "(1) The task description conveys the intended task to the language model. Following self-instruct [57], we design a set of seed instructions with placeholders and employ LLMs to generate a large number of related task descriptions and randomly select one of them during training. ",
+        "bbox": [
+            173,
+            710,
+            825,
+            752
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "(2) For conventional visual perception tasks like object detection and instance segmentation, we propose a unified output format represented as a tuple $( C , P )$ , where $C$ denotes the class index in the category set <class>, and $\\bar { P } = \\{ x _ { i } , y _ { i } \\} _ { i = 1 } ^ { N }$ represents $N$ points that locate the object. To align with the format of word tokens, both the class index and the coordinates of points $x _ { i } , y _ { i }$ are transformed into discretized tokens. Specifically, the class index is an integer starting from 0, and the continuous coordinates of the points are uniformly discretized into an integer within the range [-<range>, <range>]. For object detection and visual grounding tasks, the point number $N$ is equal to 2, representing the the top-left and bottom-right points of object’s bounding box. In the case of instance segmentation, we employ multiple $( N > 8 )$ ) points along the object boundary to represent an instance mask [61]. Other perception tasks such as pose estimation (keypoint detection) can also be formulated as language instructions in this way. ",
+        "bbox": [
+            173,
+            758,
+            825,
+            911
+        ],
+        "page_idx": 4
+    },
+    {
+        "type": "text",
+        "text": "An example of language instruction for the instance segmentation task is as follows: “Segment all the objects of category set <class> within the <range> of the image and generate a list of the format (c, x1, y1, $x 2$ , y2, ..., x8, y8). Here, c represents the index of the class label starting from $O$ , and $( x I$ , y1, x2, y2, ..., x8, y8) correspond to the offsets of boundary points of the object relative to the center point. The image is: <image>”. ",
+        "bbox": [
+            173,
+            90,
+            825,
+            161
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "3.3 Language-Guided Image Tokenizer ",
+        "text_level": 1,
+        "bbox": [
+            176,
+            171,
+            459,
+            186
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "VisionLLM considers images as a kind of foreign language and converts them into token representations. Unlike previous works [16, 52, 31] that utilize fixed-size patch embeddings to represent images, we introduce the language-guided image tokenizer to flexibly encode visual information that aligns with task-specific language prompts or instructions. ",
+        "bbox": [
+            174,
+            190,
+            825,
+            247
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "Specifically, give an image $\\mathbf { X } \\in \\mathbb { R } ^ { H \\times W \\times 3 }$ with height $H$ and width $W$ , we first feed it to the image backbones (e.g., ResNet [21]) and extract visual features $F _ { v }$ of four different scales. Additionally, we leverage a text encoder (e.g., BERT [15]) to extract the language features $F _ { l }$ from given prompts. The language features are then injected into each scale of visual features through crossattention [47], yielding multi-scale language-aware visual features, enabling the alignment of features across modalities. ",
+        "bbox": [
+            174,
+            251,
+            825,
+            337
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "Afterward, we propose to adopt a transformer-based network (e.g., Deformable DETR [73]) with $M$ random-initialized queries $Q ^ { \\dot { } } = \\{ q _ { i } \\} _ { i = 1 } ^ { M }$ to capture the high-level information of images. We build the transformer-based network on top of the multi-scale language-aware visual features to extract $M$ image tokens $T = \\{ ( e _ { i } , l _ { i } ) \\} _ { i = 1 } ^ { M }$ , each of which is represented by an embedding $e _ { i }$ and a location $l _ { i }$ denoting the semantic and positional information of the token. This design not only represents the images independent of input resolution but also extracts the visual representation that is informative with respect to the language prompts. ",
+        "bbox": [
+            174,
+            343,
+            825,
+            440
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "3.4 LLM-based Open-Ended Task Decoder ",
+        "text_level": 1,
+        "bbox": [
+            176,
+            450,
+            486,
+            465
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "We build our decoder on Alpaca [45], an LLM that is adapted from LLaMA [46], to handle various vision-related tasks with language guidance. However, Alpaca has some inherent drawbacks for vision-centric tasks, such as (1) It only has a few digit tokens (e.g., $0 { \\sim } 9$ ) in its vocabulary, which restricts its ability to locate objects by numbers; (2) It uses multiple tokens to represent the category name, resulting in an inefficient scheme in object classification; and (3) It is a causal model that is inefficient for visual perception tasks. ",
+        "bbox": [
+            173,
+            469,
+            825,
+            553
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "To tackle these issues, we expand the vocabulary of LLM with additional tokens specially designed for vision-centric tasks. First, we add a set of location tokens, denoted as $\\{ < \\mathtt { p } - 5 1 2 >$ , ..., $\\mathtt { < p 0 > }$ , ..., $\\mathsf { < p 5 1 2 > } \\}$ , where ${ \\tt A p i > }$ represents the discretized offset of $i \\in [ - 5 1 2 , 5 1 2 ]$ to the location $l _ { i }$ of the image token, and the relative value to image height or width is equal to $i / 5 1 2$ . These tokens successfully transform the object localization task from continuous variable prediction to more unified discrete bin classification. Second, we introduce semantics-agnostic classification tokens $\\{ < \\mathsf { c o } > , < \\mathsf { c } 1 > , . . . , < \\mathsf { c } 5 1 1 > \\}$ to replace category name tokens, which overcomes the inefficiency of using multiple tokens to represent categories. The mapping between category names and the classification tokens is flexibly provided in the category set <class> of language instructions, such as $\\{ \" \\mathtt { p e r s o n \" } : < \\mathtt { c 0 } >$ , \"car\": ${ < } c 1 >$ , \"black cat\": $\\mathbf { < c } 2 > , \\mathbf { \\ldots } \\}$ . This design allows our model to select the appropriate category name from the provided category set, facilitating efficient and accurate object classification. ",
+        "bbox": [
+            174,
+            559,
+            825,
+            724
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "Moreover, to address the inefficiency caused by the causal framework, we introduce outputformat-as-query decoding. We first use LLMs to parse the structural output format from the task instructions (e.g., “<cls> <x1> <y1> $< \\tt x 2 >$ $\\mathrm { < y } 2 \\mathrm { > } ^ { \\mathrm { , } \\mathrm { , } }$ for object detection, “<bos>” for image captioning), and then feed the tokens of structural output format as queries to the decoder to generate the desired output according to the queries. This simple method enables our model to not only avoid inefficient token-by-token decoding in visual perception tasks, but also keep a unified framework for vision-language tasks. ",
+        "bbox": [
+            174,
+            732,
+            485,
+            911
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "image",
+        "img_path": "images/5b925cec77c80f3ecd2208ed699c71037dba057c130a11b4919e884e7d05c4b3.jpg",
+        "image_caption": [
+            "Figure 4: Illustration of the “output-format-asquery” decoding process. $\\bf \\ddot { \\sigma } < c l s > < x 1 > < y 1 > \\tau . . . \\dot { \\sigma }$ ” denote the queries of the object’s class index and boundary points, and “<bos>” denotes the beginning of string. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            509,
+            724,
+            823,
+            828
+        ],
+        "page_idx": 5
+    },
+    {
+        "type": "text",
+        "text": "Note that, during both the training and inference phases in the object detection task, we input 100 sets of $\\mathrm { ^ { * * } { < } x l s > < x l > < y l > < x 2 > < y 2 > \" }$ to the decoder, generating 100 object predictions. Those predictions with higher confidence scores will be retained, adhering to a common practice of the object detection task. ",
+        "bbox": [
+            174,
+            90,
+            825,
+            146
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "In this way, the output of object location and classification is formulated as a foreign language, thus unifying these vision-centric tasks into the format of token classification. Therefore, both vision-language and vision-only tasks can be supervised with the cross-entropy loss like language tasks. In addition, for efficient training, we adopt the Low-Rank Adaptation (LoRA) approach [22], which allows us to train and fine-tune the models without excessive computational costs. We set the LoRA rank to 64 and use LoRA on the QKVO (Query, Key, Value, and Output) in the attention layers. It also acts as a bridge between the language and visual tokens, facilitating effective alignment between the two modalities, ensuring better task customization, and improving the convergence of the overall system. ",
+        "bbox": [
+            174,
+            154,
+            825,
+            279
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "4 Experiment ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            291,
+            303,
+            309
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "4.1 Implementation Details. ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            316,
+            379,
+            332
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "We implement two variants of VisionLLM with two image backbones, i.e., ResNet [21] and InternImage-H [51]. For the language-guided image tokenizer, we adopt BERT-Large [4] as the text encoder and Deformable DETR (D-DETR) [73] to capture high-level information. For the LLM, we employ Alpaca-7B [45], a LLaMA [46] model fine-tuned with instructions, and equip it with LoRA [22] for parameter-efficient fine-tuning. ",
+        "bbox": [
+            174,
+            343,
+            825,
+            412
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "The model is trained in two stages. In the first stage, we initialize the model with the pre-trained weights of D-DETR and BERT, and train the visual backbone and language-guided image tokenizer to produce language-aware visual features. In the second stage, we connect the image tokenizer with Alpaca-7B and introduce the unified supervision of multiple tasks. We freeze the visual backbone while freezing most parameters of the LLM except a few LoRA parameters. More details on the experimental setup can be found in Sec. B of the supplementary material. ",
+        "bbox": [
+            174,
+            417,
+            825,
+            501
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "4.2 Task-Level Customization ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            512,
+            393,
+            526
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "We first evaluate the task-level customization capability of VisionLLM. VisionLLM supports coarsegrained task customization, including visual perception tasks and visual-language tasks. Table 1 presents the evaluation results on four standard vision-centric tasks, including object detection, instance segmentation, visual grounding, and image captioning. We compare our model with taskspecific methods as well as recently-proposed vision generalist models. Note that, unless specifically mentioned, the results of our model come from a shared-parameter generalist model and switch different tasks by changing the language instructions only. Detailed instructions could be found in the supplementary material. ",
+        "bbox": [
+            174,
+            531,
+            825,
+            642
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "Object Detection. Object detection is a fundamental computer vision task that involves identifying and localizing objects of interest within an image. Our method achieves comparable or higher results to others, $4 4 . 6 \\ \\mathrm { m A P } ,$ with a ResNet-50 [21] backbone. With the same backbone i.e. ResNet-50, our method outperforms Pix2Seq [7] by $1 . 4 \\mathrm { m A P }$ , which also discretizes the output coordinates to integers. Furthermore, benefiting from the output-format-as-query framework (see Sec. 3.4), we can decode multiple predictions in parallel during inference, making our approach more efficient. Using InternImage-H [51] as the visual backbone, we obtained $6 0 . 2 \\%$ mAP, which is close to the current state-of-the-art detection-specific model [51], demonstrating the scalability of our generalist model. ",
+        "bbox": [
+            174,
+            648,
+            825,
+            760
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "Visual Grounding. Visual grounding associates textual descriptions with corresponding regions or objects within an image. Training visual grounding and object detection can potentially conflict with each other, as object detection aims to detect all the objects, while visual grounding should only localize the referred object and suppress other objects. Benefiting from our unified task instructions and the strong instruction comprehension capabilities of LLMs, our model performs both tasks effectively and achieves a result of $8 0 . 6 \\ : \\mathrm { P } @ 0 . 5$ for visual grounding. With InternImage-H as the backbone, we achieve $8 6 . 7 \\ : \\mathrm { P } @ 0 . 5$ on the validation set of RefCOCO. ",
+        "bbox": [
+            174,
+            766,
+            823,
+            863
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "text",
+        "text": "Instance Segmentation. Instance segmentation involves identifying and segmenting individual objects within an image. We employ a flexible number of points (i.e., $8 \\sim 2 4 )$ along the object boundary to represent an instance mask. Compared to mainstream models specific to instance segmentation, our model has a comparable mask $\\mathrm { { A P } _ { 5 0 } }$ ( $6 1 . 2 \\%$ with InternImage-H [51]) but relatively low mask $\\mathsf { A P } _ { 7 5 }$ . This gap could potentially arise from factors as follows: (1) We discretize the output coordinates to integers for unifying tasks, which introduces information loss; (2) Due to the memory and computational constraint, the number of points in our model is limited, which also results in a performance drop; and (3) Point-based methods typically yield lower results compared to direct mask prediction methods, such as Mask R-CNN [20]. ",
+        "bbox": [
+            176,
+            869,
+            823,
+            911
+        ],
+        "page_idx": 6
+    },
+    {
+        "type": "table",
+        "img_path": "images/808abcd8972e4247abb3e225f4a57f3de8204e7efa4a44fa9c1ce03fff5232a1.jpg",
+        "table_caption": [
+            "Table 1: Results on standard vision-centric tasks. “sep” indicates that the model is separately trained on each task. ",
+            "Table 2: Experiments of object-level and output format customization. We conduct these experiments based on VisionLLM-R50, and report the performance of box AP and mask AP on COCO minival for (a) and (b), respectively. “#Classes” and “#Points” indicate the number of classes and boundary points, respectively. “\\*” indicates that we report the mean AP of the given classes, e.g., 10 classes. "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td rowspan=\"2\">Method</td><td rowspan=\"2\">Backbone</td><td rowspan=\"2\">Open- Ended</td><td colspan=\"2\">Detection</td><td colspan=\"2\"></td><td colspan=\"2\">Instance Seg. Grounding</td><td colspan=\"2\">Captioning</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td>AP AP50 AP75 AP AP50 AP75P@0.5</td><td>BLEU-4CIDEr</td><td></td></tr><tr><td>Specialist Models</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>FasterR-CNN-FPN[40]</td><td>ResNet-50</td><td>X</td><td>40.3 61.0</td><td></td><td>）44.0</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>DETR-DC5 [6]</td><td>ResNet-50</td><td>×</td><td>43.3 63.1</td><td></td><td>45.9</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Deformable-DETR[73]</td><td>ResNet-50</td><td>X</td><td>45.7 65.0</td><td></td><td>49.1</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Mask R-CNN[20]</td><td>ResNet-50</td><td>X</td><td>41.0 61.7</td><td></td><td></td><td></td><td>44.9 37.1 58.4 40.1</td><td></td><td></td><td></td></tr><tr><td>Polar Mask [61]</td><td>ResNet-50</td><td>X</td><td>=</td><td></td><td></td><td></td><td>30.5 52.0 31.1</td><td></td><td></td><td></td></tr><tr><td>Pix2Seq[7]</td><td>ResNet-50</td><td>X</td><td>43.2 61.0 46.1</td><td></td><td></td><td></td><td>=</td><td></td><td></td><td></td></tr><tr><td>UNITER[10]</td><td>ResNet-101</td><td>X</td><td>-</td><td></td><td>=</td><td></td><td>=</td><td>81.4</td><td></td><td></td></tr><tr><td>VILLA [18]</td><td>ResNet-101</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>82.4</td><td></td><td></td></tr><tr><td>MDETR[25]</td><td>ResNet-101</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>86.8</td><td></td><td></td></tr><tr><td>BEiT-3 [53]</td><td>ViT-g</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>-</td><td></td><td>147.6</td></tr><tr><td>VL-T5 [12]</td><td>T5-B</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>1</td><td></td><td>116.5</td></tr><tr><td colspan=\"9\">GeneralistModels</td><td></td><td></td></tr><tr><td>UniTab [64]</td><td>ResNet-101</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>88.6</td><td>=</td><td>115.8</td></tr><tr><td>Uni-Perceiver[74]</td><td>ViT-B</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>-</td><td>32.0</td><td>=</td></tr><tr><td>Uni-Perceiver-MoE[72]</td><td>ViT-B</td><td>X</td><td></td><td></td><td></td><td></td><td></td><td>=</td><td>33.2</td><td>=</td></tr><tr><td>Uni-Perceiver-V2 [26]</td><td>Swin-B</td><td>X</td><td>58.6</td><td></td><td></td><td>50.6</td><td></td><td>=</td><td>35.4</td><td>116.9</td></tr><tr><td>Pix2Seq v2 [8]</td><td>ViT-B</td><td>X</td><td>46.5</td><td></td><td></td><td>38.2</td><td></td><td>1</td><td>34.9</td><td></td></tr><tr><td>VisionLLM-R50sep</td><td>ResNet-50</td><td>X</td><td>44.8 64.1 48.5 25.2 50.6 22.4</td><td></td><td></td><td></td><td></td><td>84.4</td><td>30.8</td><td>112.4</td></tr><tr><td>VisionLLM-R50</td><td>ResNet-50</td><td></td><td>44.6 64.0 4</td><td></td><td></td><td>48.1 25.1 50.0 22.4</td><td></td><td>80.6</td><td>31.0</td><td>112.5</td></tr><tr><td>VisionLLM-H</td><td>InternImage-H</td><td></td><td>60.2 79.3 65.8 30.6 61.2 27.6</td><td></td><td></td><td></td><td></td><td>86.7</td><td>32.1</td><td>114.2</td></tr></table>",
+        "bbox": [
+            178,
+            117,
+            818,
+            405
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "table",
+        "img_path": "images/1b4dee81721236cca04e12d3c14c008b64f6815bbfbdf9a6378b5442fbe2d0e9.jpg",
+        "table_caption": [
+            "(a) Object-level customization. "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td>#Classes</td><td>AP</td><td>AP50</td><td>AP75 APs</td><td>APM</td><td>APL</td></tr><tr><td>10*</td><td>48.9</td><td>72.6</td><td>51.2</td><td>31.7 47.5</td><td>67.3</td></tr><tr><td>20*</td><td>52.7</td><td>73.6</td><td>56.8</td><td>31.8 53.2</td><td>70.5</td></tr><tr><td>40*</td><td>49.3</td><td>70.7</td><td>53.2</td><td>33.1 53.6</td><td>63.8</td></tr><tr><td>80*</td><td>44.6</td><td>64.0</td><td>48.1</td><td>26.7 47.9</td><td>60.5</td></tr></table>",
+        "bbox": [
+            196,
+            505,
+            486,
+            570
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "table",
+        "img_path": "images/cc172fbc8ed4da299c668c62290a097f3cd37f6f80e58179d24f2d12e78c8549.jpg",
+        "table_caption": [
+            "(b) Output format customization. "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td>#Points</td><td>AP</td><td>AP50</td><td>AP75 APs</td><td>APM</td><td>APL</td></tr><tr><td>8</td><td>18.5</td><td>45.7</td><td>11.6</td><td>9.9 19.7</td><td>28.7</td></tr><tr><td>14</td><td>22.9</td><td>48.3</td><td>19.4</td><td>11.0 25.1</td><td>36.0</td></tr><tr><td>16</td><td>24.2</td><td>49.9</td><td>20.9</td><td>11.5 26.3</td><td>36.8</td></tr><tr><td>24</td><td>25.1</td><td>50.0</td><td>22.4</td><td>12.5 27.4</td><td>38.2</td></tr></table>",
+        "bbox": [
+            517,
+            505,
+            802,
+            570
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            173,
+            590,
+            825,
+            674
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "text",
+        "text": "Image Captioning. We also evaluate our model in a representative vision-language task, i.e. image captioning task, and report the BLEU-4 [37] and CIDEr [48] metrics. Note that we do not adopt the CIDEr optimization [41]. We can observe that VisionLLM achieves competitive performance to previous methods. With ResNet-50, we obtain a BLEU-4 score of 31.0 and a CIDEr score of 112.5. When using InternImage-H as the backbone, our model achieves a comparable BLEU-4 score of 32.1 and a CIDEr score of 114.2. These results demonstrate the effectiveness of VisionLLM in generating descriptive and contextually relevant captions for images. ",
+        "bbox": [
+            173,
+            681,
+            825,
+            779
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "text",
+        "text": "4.3 Object-Level & Output Format Customization ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            792,
+            539,
+            808
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "text",
+        "text": "Our VisionLLM not only allows for customizing the task description, but also for adjusting the target object and the output format using language instructions. Here, we evaluate our model’s fine-grained customization ability on COCO. In particular, to customize the target object, we modify the <class> in language instructions to change the model’s recognition target from 10 classes to 80 classes. Likewise, to customize the output format, we modify the number of points in language instructions to change the task output format. Table 2 shows that our method can perform well for both object-level and output format changes. ",
+        "bbox": [
+            174,
+            814,
+            825,
+            911
+        ],
+        "page_idx": 7
+    },
+    {
+        "type": "table",
+        "img_path": "images/6fd57646d70c85b07f79875e8f9de31c76e39e70f6ec64272364173884f682c5.jpg",
+        "table_caption": [
+            "(a) Effect of text encoder in the language-guided image tokenizer. "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td>W/BERT</td><td>Freeze</td><td>COCO</td><td>RefCOCO</td></tr><tr><td>1</td><td>1</td><td>44.7</td><td>48.1</td></tr><tr><td>√</td><td></td><td>44.8</td><td>84.1</td></tr><tr><td>√</td><td>√</td><td>1.3</td><td>34.3</td></tr></table>",
+        "bbox": [
+            179,
+            155,
+            419,
+            208
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "table",
+        "img_path": "images/78fbd12cb90ee238c899770a7a7020b1c89db5c0afb53829211fd5cdc2ca6483.jpg",
+        "table_caption": [
+            "Table 3: Ablation studies on language-guided image tokenizer and hyper-parameters. ",
+            "(c) Effect of the number of bins (#Bins). "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td>#Bins</td><td>AP</td></tr><tr><td>257</td><td>34.9</td></tr><tr><td>513</td><td>40.8</td></tr><tr><td>1025</td><td>44.8</td></tr><tr><td>2049</td><td>44.8</td></tr></table>",
+        "bbox": [
+            689,
+            146,
+            794,
+            212
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "table",
+        "img_path": "images/64c26c1ceb633481ec0088fbe419b41157d81baa9d5bc50f318497211445c9e9.jpg",
+        "table_caption": [
+            "(b) Effect of image tokenization method. "
+        ],
+        "table_footnote": [],
+        "table_body": "<table><tr><td>Tokenization</td><td>AP</td></tr><tr><td>Average Pooling Ours</td><td>23.1 44.8</td></tr></table>",
+        "bbox": [
+            460,
+            157,
+            622,
+            200
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "4.4 Ablation Study ",
+        "text_level": 1,
+        "bbox": [
+            173,
+            224,
+            318,
+            239
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "In this section, we analyze the effect of key components and hyper-parameters on VisionLLM. Unless otherwise specified, we use ResNet-50 [21] backbone and perform the ablation experiments for object detection tasks with random classes and task descriptions on COCO2017 [29]. ",
+        "bbox": [
+            174,
+            243,
+            825,
+            285
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Single Task vs. Multiple Tasks. We perform an ablation study to assess the impact of multi-task learning with language instructions on VisionLLM. As shown in Table 1, the single-task trained model VisionLLM- $\\cdot \\mathrm { R } 5 0 _ { \\mathrm { s e p } }$ is slightly better than the jointly trained model VisionLLM-R50 except image captioning. This is due to the multitasking conflicts that also affect previous generalist models [74, 72], and it reflects a trade-off between accuracy and generalization. ",
+        "bbox": [
+            174,
+            291,
+            825,
+            361
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Text Encoder in Language-Guided Image Tokenizer. We examine the role of text encoder (i.e., BERT) in our language-guided image tokenizer in Table 3a, where we report the results for object detection and visual grounding. The first two rows show that BERT is not essential for object detection but it is crucial for visual grounding. We also investigate the effect of freezing the text encoder during training. The last row indicates that freezing BERT hinders the alignment of vision and language modalities and thus degrades the performance for both tasks. ",
+        "bbox": [
+            174,
+            367,
+            825,
+            450
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Image Tokenization Method. As a comparison to our query-based tokenization, we employ average pooling on the feature maps from the D-DETR encoder to obtain $M$ patch embeddings, which serve as token representations for the image. Results in Table 3b indicate a clear advantage of our method. This is due to its ability to capture information from objects of various sizes in a more flexible way. ",
+        "bbox": [
+            174,
+            457,
+            825,
+            512
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Number of Localization Tokens. We vary the number of localization tokens from 257 (i.e., - $1 2 8 \\mathrm { \\sim } 1 2 8 $ ) to 2049 (i.e., - $- 1 0 2 4 { \\sim } 1 0 2 4 )$ , to investigate its impact on visual perception performance. As presented in Table 3c, the model consistently exhibits improvement as the number of localization tokens increases until it reaches a saturation point. Remarkably, a substantial performance boost is observed when the number is raised from 257 to 1025 $+ 9 . 9$ AP). These results indicate that a higher number of localization tokens enables the models to achieve finer localization abilities, thereby improving localization accuracy. ",
+        "bbox": [
+            174,
+            518,
+            825,
+            616
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "5 Conclusion ",
+        "text_level": 1,
+        "bbox": [
+            174,
+            628,
+            299,
+            645
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "In this paper, we have presented VisionLLM, a novel framework that leverages the power of large language models (LLMs) to address vision-centric tasks in an open-ended and customizable manner. We have designed unified language instruction that matches the format of language models and covers various vision-centric tasks including visual perception. We have also developed a language-guided image tokenizer and an LLM-based task decoder that can handle open-ended tasks according to the given language instructions. We have verified the effectiveness of our models on a series of tasks with different granularities, demonstrating their remarkable generality and flexibility. ",
+        "bbox": [
+            174,
+            651,
+            825,
+            750
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Broader Impact. We envision that this work will promote the fusion of visual and language tasks. In addition, since our work is built on open-source pre-trained vision foundation models and large language models, requiring low training resources, thus reducing the carbon footprint. We do not foresee obvious undesirable ethical/social impacts at this moment. ",
+        "bbox": [
+            174,
+            755,
+            825,
+            810
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "Acknowledgement ",
+        "text_level": 1,
+        "bbox": [
+            176,
+            830,
+            330,
+            847
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "The work is supported by the National Key R&D Program of China (NO. 2022ZD0161300), the National Natural Science Foundation of China (Grant No. 62376134, 61672273, 61832008, 62372223), the Shanghai Committee of Science and Technology (Grant No. 21DZ1100100), and the Fundamental Research Funds for the Central Universities (No. XJ2023000701). ",
+        "bbox": [
+            176,
+            854,
+            825,
+            910
+        ],
+        "page_idx": 8
+    },
+    {
+        "type": "text",
+        "text": "References   \n[1] Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katie Millican, Malcolm Reynolds, et al. Flamingo: a visual language model for few-shot learning. arXiv preprint arXiv:2204.14198, 2022. 1, 4   \n[2] Stanislaw Antol, Aishwarya Agrawal, Jiasen Lu, Margaret Mitchell, Dhruv Batra, C Lawrence Zitnick, and Devi Parikh. Vqa: Visual question answering. In Proceedings of the IEEE International Conference on Computer Vision, 2015. 4   \n[3] Amir Bar, Yossi Gandelsman, Trevor Darrell, Amir Globerson, and Alexei Efros. Visual prompting via image inpainting. Advances in Neural Information Processing Systems, 2022. 4   \n[4] Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is space-time attention all you need for video understanding? In International Conference on Machine Learning, 2021. 7   \n[5] Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in Neural Information Processing Systems, 2020. 3, 4   \n[6] Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detection with transformers. In European Conference on Computer Vision, 2020. 3, 4, 8   \n[7] Ting Chen, Saurabh Saxena, Lala Li, David J Fleet, and Geoffrey Hinton. Pix2seq: A language modeling framework for object detection. arXiv preprint arXiv:2109.10852, 2021. 7, 8   \n[8] Ting Chen, Saurabh Saxena, Lala Li, Tsung-Yi Lin, David J Fleet, and Geoffrey Hinton. A unified sequence interface for vision tasks. arXiv preprint arXiv:2206.07669, 2022. 4, 8   \n[9] Xinlei Chen, Hao Fang, Tsung-Yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco captions: Data collection and evaluation server. arXiv preprint arXiv:1504.00325, 2015. 4   \n[10] Yen-Chun Chen, Linjie Li, Licheng Yu, Ahmed El Kholy, Faisal Ahmed, Zhe Gan, Yu Cheng, and Jingjing Liu. Uniter: Universal image-text representation learning. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXX. Springer, 2020. 8   \n[11] Zhe Chen, Yuchen Duan, Wenhai Wang, Junjun He, Tong Lu, Jifeng Dai, and Yu Qiao. Vision transformer adapter for dense predictions. In International Conference on Learning Representations, 2023. 1   \n[12] Jaemin Cho, Jie Lei, Hao Tan, and Mohit Bansal. Unifying vision-and-language tasks via text generation. In International Conference on Machine Learning, 2021. 8   \n[13] Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416, 2022. 4   \n[14] MOSS contributors. Moss. https://github.com/OpenLMLab/MOSS, 2023. 3   \n[15] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. 6   \n[16] Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations, 2021. 6   \n[17] Yuxin Fang, Wen Wang, Binhui Xie, Quan Sun, Ledell Wu, Xinggang Wang, Tiejun Huang, Xinlong Wang, and Yue Cao. Eva: Exploring the limits of masked visual representation learning at scale. arXiv preprint arXiv:2211.07636, 2022. 1   \n[18] Zhe Gan, Yen-Chun Chen, Linjie Li, Chen Zhu, Yu Cheng, and Jingjing Liu. Large-scale adversarial training for vision-and-language representation learning. Advances in Neural Information Processing Systems, 2020. 8   \n[19] Tanmay Gupta, Amita Kamath, Aniruddha Kembhavi, and Derek Hoiem. Towards general purpose vision systems. arXiv preprint arXiv:2104.00743, 2021. 4   \n[20] Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask r-cnn. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2017. 3, 8   \n[21] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016. 6, 7, 9   \n[22] Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021. 7   \n[23] Srinivasan Iyer, Xi Victoria Lin, Ramakanth Pasunuru, Todor Mihaylov, Dániel Simig, Ping Yu, Kurt Shuster, Tianlu Wang, Qing Liu, Punit Singh Koura, et al. Opt-iml: Scaling language model instruction meta learning through the lens of generalization. arXiv preprint arXiv:2212.12017, 2022. 4   \n[24] Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Visual prompt tuning. In European Conference on Computer Vision, 2022. 1, 2   \n[25] Aishwarya Kamath, Mannat Singh, Yann LeCun, Gabriel Synnaeve, Ishan Misra, and Nicolas Carion. Mdetr-modulated detection for end-to-end multi-modal understanding. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021. 8   \n[26] Hao Li, Jinguo Zhu, Xiaohu Jiang, Xizhou Zhu, Hongsheng Li, Chun Yuan, Xiaohua Wang, Yu Qiao, Xiaogang Wang, Wenhai Wang, et al. Uni-perceiver v2: A generalist model for large-scale vision and vision-language tasks. arXiv preprint arXiv:2211.09808, 2022. 4, 8   \n[27] Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. arXiv preprint arXiv:2301.12597, 2023. 4, 5   \n[28] KunChang Li, Yinan He, Yi Wang, Yizhuo Li, Wenhai Wang, Ping Luo, Yali Wang, Limin Wang, and Yu Qiao. Videochat: Chat-centric video understanding. arXiv preprint arXiv:2305.06355, 2023. 2, 3, 5   \n[29] Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In European Conference on Computer Vision. Springer, 2014. 9   \n[30] Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. arXiv preprint arXiv:2304.08485, 2023. 4, 5   \n[31] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021. 6   \n[32] Zhaoyang Liu, Yinan He, Wenhai Wang, Weiyun Wang, Yi Wang, Shoufa Chen, Qinglong Zhang, Yang Yang, Qingyun Li, Jiashuo Yu, et al. Interngpt: Solving vision-centric tasks by interacting with chatbots beyond language. arXiv preprint arXiv:2305.05662, 2023. 2, 3, 5   \n[33] Jiasen Lu, Christopher Clark, Rowan Zellers, Roozbeh Mottaghi, and Aniruddha Kembhavi. Unified-io: A unified model for vision, language, and multi-modal tasks. arXiv preprint arXiv:2206.08916, 2022. 4   \n[34] OpenAI. Gpt-4 technical report. arXiv, 2023. 3   \n[35] TB OpenAI. Chatgpt: Optimizing language models for dialogue. OpenAI, 2022. 1, 3   \n[36] Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 2022. 3, 4   \n[37] Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, 2002. 8   \n[38] Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, et al. Improving language understanding by generative pre-training. 2018. 1, 4   \n[39] Scott Reed, Konrad Zolna, Emilio Parisotto, Sergio Gomez Colmenarejo, Alexander Novikov, Gabriel Barth-Maron, Mai Gimenez, Yury Sulsky, Jackie Kay, Jost Tobias Springenberg, et al. A generalist agent. arXiv preprint arXiv:2205.06175, 2022. 4   \n[40] Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in Neural Information Processing Systems, 2015. 8   \n[41] Steven J Rennie, Etienne Marcheret, Youssef Mroueh, Jerret Ross, and Vaibhava Goel. Self-critical sequence training for image captioning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017. 8   \n[42] Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Yueting Zhuang. Hugginggpt: Solving ai tasks with chatgpt and its friends in huggingface. arXiv preprint arXiv:2303.17580, 2023. 2, 3, 5   \n[43] Weijie Su, Xizhou Zhu, Chenxin Tao, Lewei Lu, Bin Li, Gao Huang, Yu Qiao, Xiaogang Wang, Jie Zhou, and Jifeng Dai. Towards all-in-one pre-training via maximizing multi-modal mutual information. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023. 1   \n[44] Chenxin Tao, Xizhou Zhu, Gao Huang, Yu Qiao, Xiaogang Wang, and Jifeng Dai. Siamese image modeling for self-supervised vision representation learning. arXiv preprint arXiv:2206.01204, 2022. 1   \n[45] Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B Hashimoto. Alpaca: A strong, replicable instruction-following model. Stanford Center for Research on Foundation Models. https://crfm. stanford. edu/2023/03/13/alpaca. html, 2023. 3, 6, 7   \n[46] Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. 3, 6, 7   \n[47] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in Neural Information Processing Systems, 30, 2017. 6   \n[48] Ramakrishna Vedantam, C Lawrence Zitnick, and Devi Parikh. Cider: Consensus-based image description evaluation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015. 8   \n[49] Jianfeng Wang, Zhengyuan Yang, Xiaowei Hu, Linjie Li, Kevin Lin, Zhe Gan, Zicheng Liu, Ce Liu, and Lijuan Wang. Git: A generative image-to-text transformer for vision and language. arXiv preprint arXiv:2205.14100, 2022. 1, 4   \n[50] Peng Wang, An Yang, Rui Men, Junyang Lin, Shuai Bai, Zhikang Li, Jianxin Ma, Chang Zhou, Jingren Zhou, and Hongxia Yang. Unifying architectures, tasks, and modalities through a simple sequence-tosequence learning framework. arXiv preprint arXiv:2202.03052, 2022. 1, 4   \n[51] Wenhai Wang, Jifeng Dai, Zhe Chen, Zhenhang Huang, Zhiqi Li, Xizhou Zhu, Xiaowei Hu, Tong Lu, Lewei Lu, Hongsheng Li, et al. Internimage: Exploring large-scale vision foundation models with deformable convolutions. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023. 1, 2, 7, 8   \n[52] Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. Pvt v2: Improved baselines with pyramid vision transformer. Computational Visual Media, 8(3):415–424, 2022. 6   \n[53] Wenhui Wang, Hangbo Bao, Li Dong, Johan Bjorck, Zhiliang Peng, Qiang Liu, Kriti Aggarwal, Owais Khan Mohammed, Saksham Singhal, Subhojit Som, et al. Image as a foreign language: Beit pretraining for all vision and vision-language tasks. arXiv preprint arXiv:2208.10442, 2022. 1, 2, 8   \n[54] Xinlong Wang, Wen Wang, Yue Cao, Chunhua Shen, and Tiejun Huang. Images speak in images: A generalist painter for in-context visual learning. arXiv preprint arXiv:2212.02499, 2022. 1, 2   \n[55] Xinlong Wang, Wen Wang, Yue Cao, Chunhua Shen, and Tiejun Huang. Images speak in images: A generalist painter for in-context visual learning. arXiv preprint arXiv:2212.02499, 2022. 4   \n[56] Xinlong Wang, Xiaosong Zhang, Yue Cao, Wen Wang, Chunhua Shen, and Tiejun Huang. Seggpt: Segmenting everything in context. arXiv preprint arXiv:2304.03284, 2023. 2, 4   \n[57] Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A Smith, Daniel Khashabi, and Hannaneh Hajishirzi. Self-instruct: Aligning language model with self generated instructions. arXiv preprint arXiv:2212.10560, 2022. 4, 5   \n[58] Yizhong Wang, Swaroop Mishra, Pegah Alipoormolabashi, Yeganeh Kordi, Amirreza Mirzaei, Anjana Arunkumar, Arjun Ashok, Arut Selvan Dhanasekaran, Atharva Naik, David Stap, et al. Benchmarking generalization via in-context instructions on $1 { , } 6 0 0 { + }$ language tasks. arXiv preprint arXiv:2204.07705, 2022. 4   \n[59] Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. arXiv preprint arXiv:2109.01652, 2021. 4   \n[60] Chenfei Wu, Shengming Yin, Weizhen Qi, Xiaodong Wang, Zecheng Tang, and Nan Duan. Visual chatgpt: Talking, drawing and editing with visual foundation models. arXiv preprint arXiv:2303.04671, 2023. 2, 3, 5   \n[61] Enze Xie, Peize Sun, Xiaoge Song, Wenhai Wang, Xuebo Liu, Ding Liang, Chunhua Shen, and Ping Luo. Polarmask: Single shot instance segmentation with polar representation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020. 5, 8   \n[62] Bin Yan, Yi Jiang, Jiannan Wu, Dong Wang, Ping Luo, Zehuan Yuan, and Huchuan Lu. Universal instance perception as object discovery and retrieval. arXiv preprint arXiv:2303.06674, 2023. 4   \n[63] Zhengyuan Yang, Zhe Gan, Jianfeng Wang, Xiaowei Hu, Faisal Ahmed, Zicheng Liu, Yumao Lu, and Lijuan Wang. Unitab: Unifying text and box outputs for grounded vision-language modeling. In European Conference on Computer Vision, pages 521–539. Springer, 2022. 4   \n[64] Zhengyuan Yang, Zhe Gan, Jianfeng Wang, Xiaowei Hu, Faisal Ahmed, Zicheng Liu, Yumao Lu, and Lijuan Wang. Unitab: Unifying text and box outputs for grounded vision-language modeling. In European Conference on Computer Vision, 2022. 8   \n[65] Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Ehsan Azarnasab, Faisal Ahmed, Zicheng Liu, Ce Liu, Michael Zeng, and Lijuan Wang. Mm-react: Prompting chatgpt for multimodal reasoning and action. arXiv preprint arXiv:2303.11381, 2023. 2, 3, 5   \n[66] Yuan Yao, Ao Zhang, Zhengyan Zhang, Zhiyuan Liu, Tat-Seng Chua, and Maosong Sun. Cpt: Colorful prompt tuning for pre-trained vision-language models. arXiv preprint arXiv:2109.11797, 2021. 1   \n[67] Yuhang Zang, Wei Li, Kaiyang Zhou, Chen Huang, and Chen Change Loy. Unified vision and language prompt learning. arXiv preprint arXiv:2210.07225, 2022. 1   \n[68] Aohan Zeng, Xiao Liu, Zhengxiao Du, Zihan Wang, Hanyu Lai, Ming Ding, Zhuoyi Yang, Yifan Xu, Wendi Zheng, Xiao Xia, et al. Glm-130b: An open bilingual pre-trained model. arXiv preprint arXiv:2210.02414, 2022. 3   \n[69] Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, et al. Opt: Open pre-trained transformer language models. arXiv preprint arXiv:2205.01068, 2022. 3   \n[70] Yuanhan Zhang, Kaiyang Zhou, and Ziwei Liu. Neural prompt search. arXiv preprint arXiv:2206.04673, 2022. 1   \n[71] Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. Minigpt-4: Enhancing visionlanguage understanding with advanced large language models. arXiv preprint arXiv:2304.10592, 2023. 4   \n[72] Jinguo Zhu, Xizhou Zhu, Wenhai Wang, Xiaohua Wang, Hongsheng Li, Xiaogang Wang, and Jifeng Dai. Uni-perceiver-moe: Learning sparse generalist models with conditional moes. arXiv preprint arXiv:2206.04674, 2022. 1, 4, 8, 9   \n[73] Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable detr: Deformable transformers for end-to-end object detection. In International Conference on Learning Representations, 2021. 3, 6, 7, 8   \n[74] Xizhou Zhu, Jinguo Zhu, Hao Li, Xiaoshi Wu, Hongsheng Li, Xiaohua Wang, and Jifeng Dai. Uniperceiver: Pre-training unified architecture for generic perception for zero-shot and few-shot tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. 2, 4, 5, 8, 9 ",
+        "bbox": [
+            169,
+            64,
+            826,
+            926
+        ],
+        "page_idx": 9
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            171,
+            35,
+            828,
+            921
+        ],
+        "page_idx": 10
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            171,
+            65,
+            828,
+            916
+        ],
+        "page_idx": 11
+    },
+    {
+        "type": "text",
+        "text": "",
+        "bbox": [
+            171,
+            92,
+            828,
+            801
+        ],
+        "page_idx": 12
+    }
+]

parse/train/3hGNqpI4WS/3hGNqpI4WS_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84ab204c641eb18ace53380e76b264ce0e205e9f15cd51447b765699056e38f1
+size 4825977

parse/train/3hGNqpI4WS/3hGNqpI4WS_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29c16ea90e652980ef2ad0032385f67bc1b177e02014ca24fea4d05e24c2bea9
+size 4514375

parse/train/3hGNqpI4WS/3hGNqpI4WS_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4faa279a64ed9fa563c8e06c6614157a9ee84bf0005317554121b6c9fe962bde
+size 4823883

parse/train/6Tm1mposlrM/6Tm1mposlrM_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53c600fb02dac1d03abb70922b9ff81b33f4d256276026adcd96b3f157db188f
+size 1106068

parse/train/6Tm1mposlrM/6Tm1mposlrM_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be17fe20455f0f2ca84c62102e7b0e4cf33b5b99e7351d2fb97ff39a969c651
+size 816224

parse/train/6Tm1mposlrM/6Tm1mposlrM_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fabae56dab7824f7a5bbe0618ee042f61580760bd5b8f05f3d03bede83d97e4a
+size 1110570

parse/train/A7pvvrlv68/A7pvvrlv68_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6ac96cffbbd9dc7bf28dd5cf1d330554eecd9de5a2946a687a7dfcfb3a9a142
+size 767095

parse/train/A7pvvrlv68/A7pvvrlv68_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f44158a10f7e73d4fc45d03efb66b40539510c23665d74c19041a0a65dc31e6
+size 587599

parse/train/A7pvvrlv68/A7pvvrlv68_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de06790e47f9104f11815bc88d52c1c3ddc14c6930e0c7804432ad736cac6c28
+size 789334

parse/train/B13njo1R-/B13njo1R-_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b611aea797f90ce9175020b74df4745fc26800a965934cb55ef3fa95e0194adb
+size 4088521

parse/train/B13njo1R-/B13njo1R-_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9f7eb33cd71872cd24751554a486b4cf62edd2b736495cfc8c7993c4b9f1a18
+size 3894763

parse/train/B13njo1R-/B13njo1R-_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba071d787c642d9f70d9e172b2b9a87b64ec9854811f060bce59628e0a9dbc88
+size 4089765

parse/train/B1xGGTEtDH/B1xGGTEtDH_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a71157fb18d0c9f221e811bc7ac01ea3e7c389ea5f629f8a4ecde1a71bc0996
+size 677960

parse/train/B1xGGTEtDH/B1xGGTEtDH_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b0ae786eba0290c4e6c0009deaef259be36b8d4776a8ad34cf91b4920e0557b
+size 344471

parse/train/B1xGGTEtDH/B1xGGTEtDH_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0b73fecb4eb7654127b6cd7985fb1950b13fcd86d3ecdb74db438cd507b1cf8
+size 698029

parse/train/B1xSperKvH/B1xSperKvH_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68dad34e60730d4f62f5c3645cabd02c74bf208ebc7786e3e68bd77d76066b93
+size 1647749

parse/train/B1xSperKvH/B1xSperKvH_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd5d0b28cf41c3ee1f4133dbbd7d9c3dcfeea4ff6df16e5d8a3e950686f14dba
+size 1509582

parse/train/B1xSperKvH/B1xSperKvH_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6c4f1cc2600af68018db604f9fb331838f5b608947d81d49bdf0a0d144a45cf
+size 1655984

parse/train/BkJsCIcgl/BkJsCIcgl_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e34e533f6b1803735fdd35461db17ff4958e340331cd0f68cb9d24a54b7e94a
+size 1928850

parse/train/BkJsCIcgl/BkJsCIcgl_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:293a063fe887071ca9d0600f9860a6164c86a85d778f3448fff8f1f4f9cfa670
+size 1778201

parse/train/BkJsCIcgl/BkJsCIcgl_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e00297979a00afe18e11ce57b344085f6740511c81e7c0399bf5db38ef85dca
+size 1938611

parse/train/BkSDMA36Z/BkSDMA36Z_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d01febbe214ee8726e8af51a6d04fa12e041a2f180a38b55f0cffb3fdd1a6de3
+size 1268945

parse/train/BkSDMA36Z/BkSDMA36Z_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b72a1d17fc5d722179856a15718f296b3a9ee1641a933f5b949261042d56542c
+size 964642

parse/train/BkSDMA36Z/BkSDMA36Z_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba8bfca29924642093e33078a71981fa5ea95e79b85811354a4587b2785efd73
+size 1269987

parse/train/ByC7ww9le/ByC7ww9le_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb52808f2785964f1384c615b3400b1cbe7eb737ba104d9ff2ef67c9f1ba1183
+size 2166800

parse/train/ByC7ww9le/ByC7ww9le_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af0083cfa809146f714e31331af6f31607100cc7f90062d77c8385efc5d75ed7
+size 1907166

parse/train/ByC7ww9le/ByC7ww9le_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71e6d4afe244363f86de483df6d0124366ad5304e9409325be497340bcdb9191
+size 2168824

parse/train/ByJHuTgA-/ByJHuTgA-_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f053faea4460cca465e355b8e79b3f53b75fe2a535642f9c7bce82f936e783ac
+size 1002208

parse/train/ByJHuTgA-/ByJHuTgA-_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1497bdc06c50af53c31abd4761ef6aeeb24ba44807f72798c088fdba0758b4d6
+size 917796

parse/train/ByJHuTgA-/ByJHuTgA-_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:deb0fd5e63b79413e60c3e5fc6e0ce1055a91c77a178555178fe97af87c7414f
+size 998559

parse/train/ByMVTsR5KQ/ByMVTsR5KQ_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca2b767413b0b2dbfe45a23e952902be6be6e59053fcdec7a13aacf3ed5ed9e7
+size 2538564

parse/train/ByMVTsR5KQ/ByMVTsR5KQ_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a476d77fac0ff1db859dcaf882a61e77ad8cd3fb26f072c8146018544378f83
+size 2366000

parse/train/ByMVTsR5KQ/ByMVTsR5KQ_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da2b4d9f796e3f06ceb573bc2122856e58ff38b081b8e618a3c693c4deefa3d3
+size 2537632

parse/train/ByeSdsC9Km/ByeSdsC9Km_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d99066b456b51fffb45aefc79912f28d9ba7712e4fea6e207eee04d9f73b59fa
+size 1608530

parse/train/ByeSdsC9Km/ByeSdsC9Km_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb2176b76e32d749caa46607aebb02eb60c9839f7e708b6567017f5da2fbf799
+size 1505719

parse/train/ByeSdsC9Km/ByeSdsC9Km_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae2d6011c17be9976eaaf7eaafe2eababb9bf8f4ef37143b79f2f51b606f2a75
+size 1608397

parse/train/ByexElSYDr/ByexElSYDr_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5efd62920d23d41e3e831d7a7e7a0db0dd07858896ae036ae6d372c715aa29f
+size 2382708

parse/train/ByexElSYDr/ByexElSYDr_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4e6695b64ec1cb3c1684f5d29e4646ea9be8ff18a45c36df177ea6d40d5483e
+size 1949481

parse/train/ByexElSYDr/ByexElSYDr_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69b4b578e0221fb194595ff2a5d30eab9e9c9659daab6da72e1ac7c5a1477cdd
+size 2408036

parse/train/Byl8hhNYPS/Byl8hhNYPS_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc0e417f69af6e026ef9a64af6e6b034b4b81837469803d8983d67990ce3bdc3
+size 16399114

parse/train/Byl8hhNYPS/Byl8hhNYPS_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85b3d3d55c434f5b9d8ab7e95338d4d6f21979d927ed3d67bc37b96e36468050
+size 16268109

parse/train/Byl8hhNYPS/Byl8hhNYPS_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:636e11c9c3f9b14f4be61f347fa07cf72eed43bd41a8b897fbe6d34555002126
+size 16399844

parse/train/CGFN_nV1ql/CGFN_nV1ql_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bce08964fb77a2b0fa3bbe40bf37371f9f65c608aecb0451f712c771a523876e
+size 848514

parse/train/CGFN_nV1ql/CGFN_nV1ql_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c179b2c43c0bd79c37eb83ea8df5bd30e7c0a75913f52b7959ad1ed9672f7f51
+size 716128

parse/train/CGFN_nV1ql/CGFN_nV1ql_span.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9da3172e383b79cef9ee3c1cbbe17a6fd7f88fc6c1ea1ab11921b9aad0c832d3
+size 846046

parse/train/H1gL-2A9Ym/H1gL-2A9Ym_layout.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc6ef75de8bfd984f0040e3625c49030d3779cfacec2c8daff8dcafb6192d599
+size 900580

parse/train/H1gL-2A9Ym/H1gL-2A9Ym_origin.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56f7b856d707a78b87441fdf65d9cfab865616b0af5c4103e84cb96d2e2318ed
+size 446784